search this blog

Friday, May 6, 2016

Villabruna cluster =/= Near Eastern migrants


I've been running a lot of Treemix analyses with the samples from the recent Qiaomei Fu et al. paper. And the impression I'm getting is that the authors missed the elephant in the room, the one with R1b painted on its big butt.

Now, it's true that Treemix output can't be used as unambiguous evidence in support of complex models. That's because in the absence of key samples the algorithm can get exceedingly creative in modeling the available data, sometimes to such extremes that the results might seem absurd.

However, when something keeps showing up again and again, even when using somewhat different samples and marker sets, and makes sense in the context of haploid and archaeological data, then at the very least it deserves serious consideration.

Here's a nice Treemix series that more or less captures the meat and potatoes of my many Treemix experiments with the Qiaomei Fu et al. dataset. For the archaeological contexts and other details about these ancient samples see here.








Below is my interpretation of the results:

- Villabruna is a sister clade of the earlier European Vestonice clade, but with significant input from an AfontovaGora3-related North Eurasian population, perhaps one that was living north of the Black Sea after the Kostenki people went the way of the dodo

- Hence, the R1b lineage carried by Villabruna I9030, the individual in this Treemix series, probably comes from the Eurasian steppe

- Kotias, a Caucasus Hunter-Gatherer, is in large part derived from the same North Eurasian population, hence the close relationship between Villabruna and Caucasus Hunter-Gatherers

- Villabruna and/or closely related foragers contributed significant ancestry to Neolithic Anatolians, and thus, indirectly, possibly to all extant Near Eastern and even many African populations

- Kotias is also closely related to Neolithic Anatolians, but probably mostly via a more basal population, perhaps the so called Basal Eurasians, native to the Near East prior to the Villabruna and/or related gene flow across the Near East and parts of Africa

- Present-day East Asians might be ancient hybrids with admixture from the same or very similar North Eurasian population, although as per the above mentioned quirks of Treemix, it's possible that the North Eurasians that contributed ancestry to Villabruna, Caucasus foragers and Eurasian steppe populations were in fact partly East Asian

So basically what I'm seeing are back migrations from Europe and the Eurasian steppe or Siberia to the Near East soon after the Ice Age. Yes, from Europe, although I admit that things can get fuzzy here.

That's because at the time much of the Aegean Sea was dry land, and thus there was no geographic barrier between the European Balkans and Asian Anatolia. The two regions, which might seem very distinct to us today, were basically one. So when I say Europe, I actually mean the ancient landmass now divided between the Balkans and Anatolia.

Or not? Are there any D-stats that we can run to either confirm or debunk my Treemix-based hypothesis? Feel free to post your proposals in the comments and I'll try and run them as soon as possible.

Update 07/05/2016: The plot thickens somewhat. I added MA1 to the line up, and now Villabruna shows minor Kotias-related ancestry. In other words, probably something from the Caucasus. The full series is available in a zip file here.


However, although interesting, this result doesn't change much. In fact, it adds more weight to the argument that Villabruna inherited his Y-chromosome from steppe foragers. That's because if this migration edge was a reflection of gene flow from Anatolia to Europe, I'd expect it to run from the Anatolia Neolithic branch, rather than the Kotias branch.

See also...

Following the mammoth herds?

227 comments:

1 – 200 of 227   Newer›   Newest»
Shaikorth said...

Could you make that tree with 2-7 migration edges and Papuans added?

Krefter said...

Interesting stuff. I have a feeling all non-U mtDNA and non-IJ/C1/R Y DNA in West Eurasia is derived from Basal Eurasian.

Rob said...

That is my reading of their Suppl material too; as I mentioned on your earlier thread; but it's good to see the tree mix figures the same (pending further aDNA)

One thing- I don't think the Aegean basin was land even during the LGM, rather the land bridge was the Bosoprus and a much smaller Black Sea (a "Black Lake")

ryukendo kendow said...

AfontovaGora3 Mbuti Villabruna ElMiron
AfontovaGora3 Mbuti Villabruna GoyetQ116-1
AfontovaGora3 Mbuti Villabruna Kostenki14
AfontovaGora3 Mbuti Villabruna Vestonice
AfontovaGora3 Mbuti ElMiron GoyetQ116-1
AfontovaGora3 Mbuti ElMiron Kostenki14
AfontovaGora3 Mbuti ElMiron Vestonice
AfontovaGora3 Mbuti GoyetQ116-1 Kostenki14
AfontovaGora3 MA1 GoyetQ116-1 Vestonice
Mbuti AfantovaGora3 Kostenki14 Vestonice
Mbuti MA1 Villabruna ElMiron
Mbuti MA1 Villabruna GoyetQ116-1
Mbuti MA1 Villabruna Kostenki14
Mbuti MA1 Villabruna Vestonice
Mbuti MA1 ElMiron GoyetQ116-1
Mbuti MA1 ElMiron Kostenki14
Mbuti MA1 ElMiron Vestonice
Mbuti MA1 GoyetQ116-1 Kostenki14
Mbuti MA1 GoyetQ116-1 Vestonice
Mbuti MA1 Kostenki14 Vestonice

Looks like the whole Villabruna-NAms thing from visual inspection of f3s may have something to it after all.

ryukendo kendow said...

@David
Another thing, some people have already mentioned that the table where they show Mal'ta as an outgroup to all pre-neolithic Europeans onevertheless has a bunch of Z=2o and above. But yhe stats are highly erratic and may be lik3 the E Asian affinity in being spread very unevenly. You might want to check that too.

Davidski said...

@Shaikorth

https://drive.google.com/file/d/0B9o3EYTdM8lQTFpGRXNHOVFjVnM/view?usp=sharing

But FYI for everyone, I can't run anymore Treemix requests. They don't take long to run, but more than a few minutes, which is too long.

Davidski said...

@rk

AfontovaGora3 Mbuti Villabruna ElMiron 0.0144 2.382 187772
AfontovaGora3 Mbuti Villabruna GoyetQ116-1 0.0179 2.748 195587
AfontovaGora3 Mbuti Villabruna Kostenki14 0.0281 4.465 218311
AfontovaGora3 Mbuti Villabruna Vestonice 0.0199 3.844 213665
AfontovaGora3 Mbuti ElMiron GoyetQ116-1 0.0031 0.38 165387
AfontovaGora3 Mbuti ElMiron Kostenki14 0.0154 1.963 174132
AfontovaGora3 Mbuti ElMiron Vestonice 0.0081 1.169 179675
AfontovaGora3 Mbuti GoyetQ116-1 Kostenki14 0.0084 1 182192
AfontovaGora3 MA1 GoyetQ116-1 Vestonice -0.0082 -0.822 155129
Mbuti AfontovaGora3 Kostenki14 Vestonice 0.0085 1.217 199942
Mbuti MA1 Villabruna ElMiron -0.0129 -2.561 427984
Mbuti MA1 Villabruna GoyetQ116-1 -0.002 -0.326 476539
Mbuti MA1 Villabruna Kostenki14 -0.0124 -2.339 644492
Mbuti MA1 Villabruna Vestonice -0.0061 -1.388 531468
Mbuti MA1 ElMiron GoyetQ116-1 0.0126 1.768 350794
Mbuti MA1 ElMiron Kostenki14 -0.0004 -0.059 408949
Mbuti MA1 ElMiron Vestonice 0.0064 1.104 389467
Mbuti MA1 GoyetQ116-1 Kostenki14 -0.0101 -1.439 456846
Mbuti MA1 GoyetQ116-1 Vestonice -0.0015 -0.237 418901
Mbuti MA1 Kostenki14 Vestonice 0.0076 1.324 511090

Karl_K said...

@Krefter

"Interesting stuff. I have a feeling all non-U mtDNA and non-IJ/C1/R Y DNA in West Eurasia is derived from Basal Eurasian."

It is interesting that there are not any super early diverging haplogroup branches associated with the Basal Eurasian autosomal component.



epoch2013 said...

Greek mesolithic HG's were mtDNA K1c.

http://eurogenes.blogspot.nl/2015/11/first-neolithic-genomes-from-greece.html

That may be a sign that they had more affinity with Middle-East. However, all Villabruna cluster samples have some sort of mtDNA U5. The distance Greece-Adriatic isn't unsurmountable 14.000 year ago.

OTOH samples of El Miron cluster also show elevated Middle-Eastern affinity.

Matt said...

That's quite a deeply divergent fork of the ANE clade, really a trifurcation of the Kotias-Afontagova-OtherVillabruna set (or a quadfuraction, including what is going to ENA). Assuming the structure works, is the OtherVillabruna differentiation even necessarily going to happen on the steppe? I'm not sure it would have to. I'm also not totally convinced that Treemix is very good for directionality of these early divergences, in general.

Ah, shame you're not running any more treemix. I think to test the structure, running again introducing Malta1, Karelia_HG and Motala (or just Malta1) would be useful. For Malta1 because it's slightly less related to Villabruna than AfontovaGora3 and KareliaHG somewhat more to Villabruna.

(I'd also have liked to see it with Oase1 and Neanderthal if possible).

Davidski said...

I'd run out of markers if I put all of those genomes into the analysis.

I don't think MA1 is all that useful as a proxy for the North Eurasian admixture in West Eurasia. Not sure why, maybe it's too old or just an extinct side branch?

AfontovaGora3 is much better. Pity it's not a high coverage genome.

epoch2013 said...

I have been thinking that HG's were essentially different from pastoralists. The idea is this: While latter brought their animals to greener pastures, and thus could consider crossing bounderies the animals would never consider, the former *followed* the animals to greener pastures. That would mean Yamnaya people migrated according to their own planning, whereas HG's migrated following existing animal migratory trajectories. Also, LBK migrated to Löss grounds and sowed cereals, rather than stay in areas with wild cereals.

So maybe you can't just say this or that HG migrated westward without making a biological case for it.

Davidski said...

Very interesting to compare AfontovaGora3 to MA1 here. And keep in mind MA1 offers many more markers.

Mbuti AfontovaGora3 Villabruna Vestonice -0.0199 -3.844 213665
Mbuti AfontovaGora3 Villabruna Kostenki14 -0.0281 -4.465 218311

Mbuti MA1 Villabruna Vestonice -0.0061 -1.388 531468
Mbuti MA1 Villabruna Kostenki14 -0.0124 -2.339 644492

Aram said...

Something or somebody made Paleo-Europeans and Paleo-Near Easterns closer to each other. Maybe Paleo-South Asians had a same influence?
History repeats himself, what happened in Bronze Age already happened in Paleolithic.

Alberto said...

AfontoveGora3 seems to have some Villabruna related ancestry that is missing in MA1. But AG3 already postdates ElMiron in Iberia, so it's difficult to establish a AG3 -> ElMiron trajectory. Especially when, just like in Europe, that Villabruna ancestry looks like a new arrival to Siberia.

Davidski said...

The D-stats seem to be picking up North Eurasian gene flow into Europe starting during the ElMiron period.

Mbuti AfontovaGora3 ElMiron Vestonice -0.0081 -1.169 179675
Mbuti AfontovaGora3 Villabruna ElMiron -0.0144 -2.382 187772
Mbuti AfontovaGora3 Villabruna Vestonice -0.0199 -3.844 213665

There's no way this is European or Near Eastern admixture in Siberia when AG3 was alive. The ice sheets hadn't even melted properly back then in Siberia. It was still the Ice Age in Afontova Gora.

Olympus Mons said...

Davidski
In this treemix kotias is generically the same as satsurblia, right?

The reason I keep on asking is because I am pretty sure there will be a lot of r1b not later then kotias in the southern caucasus. Buckloads by late 8m bc.

Then is just my pet theory. With a lot more BE they became shulaveri and later bell beaker by a southern route.
Everybody needs a pet right?
Yours clearly is a belly buttom in and around volga. Would dnieper basin still work for you as the spring of r1b, right?

Alberto said...

@Davidski

ElMiron predates AfontovaGora3. And the Villabruna-like ancestry is missing in MA1. Now that we have more West Eurasian samples, MA1 is not even longer considered a West Eurasian branch, but something that branched off earlier. And no MA1 related ancestry is present in Villabruna.

There's no way this is European or Near Eastern admixture in Siberia when AG3 was alive. The ice sheets hadn't even melted properly back then in Siberia. It was still the Ice Age in Afontova Gora.

Whatever barrier could be for gene flow, it had to be both ways. Though apparently it was not unsurmountable (whichever way the gene flow went).

Davidski said...

Yeah, Kotias is just a younger version of Satsurblia.

But I'd say the problem with your pet theory is the lack of any R1b in the Near East prior to the Middle Bronze Age.

Both Kotias and Satsurblia belong to hg J. There's also evidence of CHG gene flow into the western Anatolian Neolithic farmers, along with hg J, but no R1b.

So while it's not possible to claim that there was no R1b in the Near East and Caucasus before the Bronze Age, to say there was buckloads is being too enthusiastic at this stage IMO.

And I don't know where R1b comes from, but it's probably somewhere north of the Caucasus.

Davidski said...

Alberto,

ElMiron predates AfontovaGora3, but not the ancestors and relatives of AfontovaGora3 who lived when ElMiron was alive.

Indeed, I'm not suggesting that the AfontovaGora3 population migrated to Europe. I think that their close relatives did from some place west of Afontova Gora, from which it was easier to get to Europe because it wasn't surrounded by ice sheets.

Alberto said...

Yes, exactly, one population contributed to both ElMiron (around 40%) and AfontovaGora3 (some 10%?). That's the correct way to put it. A population represented by Villabruna.

We don't know yet where this population came from, but we know that ancient Near Easterners (from Anatolia, but probably true for the Levant too) had a large amount of this kind of ancestry, and that the Caucasus had a much smaller amount of this kind of ancestry (probably true for everywhere east of the Fertile Crescent).

In AdmixtureGraph, Villabruna picks up some 1-2% ancestry from something very basal, but the authors hypothesize that it might be related to Villabruna being non-UDG-treated. Fair enough. But it doesn't pick any MA1 related ancestry (and AG3 belongs to the Malta1 cluster). So the part concerning Villabruna -> AG3 gene flow and not the other way around looks clear.

Davidski said...

Villabruna has North Eurasian admixture. The problem is that this North Eurasian admixture is not well represented by MA1.

Of course, the problem is solved with AG3.

Davidski said...

So you reckon AG3 has Villabruna admixture? How come it doesn't show up here?

Mbuti Villabruna AfontovaGora3 MA1 0.002 0.311 170325

And yet it does show up in Karelia_HG.

Mbuti Villabruna AfontovaGora3 Karelia_HG 0.0669 11.729 224962
Mbuti Villabruna MA1 Karelia_HG 0.0633 13.091 618065

epoch2013 said...

Siberia was relatively ice free during LGM.

http://www.sonoma.edu/users/f/freidel/global/fig%209.4%20ice%20sheets.JPG

I had a discussion with Rob before, and we both agreed that the information is sketchy, but we could take a look at the distribution of Woolly Mammoths.

http://www.nature.com/ncomms/journal/v3/n6/fig_tab/ncomms1881_F2.html

This too is sketchy, but it does show a line with Mammoth finds from Lake Baikal to China where Mammoth finds and evidence for humans coincide. If this is any clue contact between ANE and UP Europa has to be before 15.000 year, or possibly later. The map does not clearly show Europe, but the demise of the Mammoth in Western Siberia looks from around the time of the Gravettian demise and the population turnover.

epoch2013 said...

Ah, a map with Mammoth finds in Europe: https://en.wikipedia.org/wiki/Woolly_mammoth#/media/File:Woolly_Mammoth_Climatic_Suitability_-_Nogu%C3%A9s-Bravo_2008.png

Alberto said...

So you reckon AG3 has Villabruna admixture?

That's what it seemed based on the stats you posted above. I would have liked the paper to take a better look at AG3, but they mostly ignored it.

How come it doesn't show up here?

Mbuti Villabruna AfontovaGora3 MA1 0.002 0.311 170325


Hhmmm... You tell me. So there is no gene flow between Villabruna and AG3, but rather gene flow between GoyetQ116-1 (and/or other old groups) and MA1? I guess we need to check more stats to really know that. These come to mind, but probably RK can come up with the best strategy here.

Mbuti GoyetQ116-1 MA1 AfontovaGora3
Mbuti Kostenki14 MA1 AfontovaGora3

Matt said...

Davidski: I don't think MA1 is all that useful as a proxy for the North Eurasian admixture in West Eurasia. Not sure why, maybe it's too old or just an extinct side branch?

AfontovaGora3 is much better. Pity it's not a high coverage genome.


Statistically, AfontovaGora3 seems better than Malta1 for Karelia and Native Amerians, as:

Malta1 AfontovaGora3 Karelia Mbuti -0.0055 -7.5 444390
Malta1 AfontovaGora3 Karitiana Mbuti -0.0030 -5.3 510778

But not any better for Villabruna than Malta1:

Malta1 AfontovaGora3 ElMiron Mbuti 0.0001 0.1 354861
Malta1 AfontovaGora3 Villabruna Mbuti -0.0004 -0.6 459597
Malta1 AfontovaGora3 Ranchot88 Mbuti -0.0001 -0.2 223016
Malta1 AfontovaGora3 Loschbour Mbuti 0.0004 0.7 505247
Malta1 AfontovaGora3 LaBrana1 Mbuti -0.0008 -1.3 489381
Malta1 AfontovaGora3 Hungarian.KO1 Mbuti -0.0005 -0.8 354548

It seems like Malta1 and AfontovaGora3 are a clade with respect to Villabruna, so any contributor to Villabruna from ANE would have to diverge before / simultaneously to Malta1 and AfontovaGora3 lineages diverging from one another. (and that would contraindicate a contribution from an ancestor of Villabruna to AfontovaGora3 to the exclusion of Malta).

Ancestry from something that diverged from the common ancestor of Satsurblia cluster, AfontovaGora3 and Malta1 would help explain why Villabruna is closer to all three (compared to the populations who came before Villabruna in Europe)?

Davidski said...

Alberto,

No, Villabruna has significantly inflated AG3-related ancestry compared to Vestonice. This looks like a pulse of admixture from the east after Vestonice.

Mbuti AfontovaGora3 Villabruna Vestonice -0.0199 -3.844 213665

But AG3 doesn't show inflated Villabruna/El Miron-related ancestry compared to MA1. In other words, there was no pulse of admixture from the west into Siberia just before or during AG3.

Mbuti Villabruna AfontovaGora3 MA1 0.002 0.311 170325
Mbuti ElMiron AfontovaGora3 MA1 0.0093 1.043 138325

So like I said, Villabruna has AG3-related North Eurasian admixture, as per the Treemix graphs. But AG3 doesn't have any Villabruna admixture, unless MA1 has it as well.

Your D-stats show that there may have been admixture from the west into Siberia before or during MA1, but this has no implications for the relationship between AG3 and Villabruna.

Mbuti GoyetQ116-1 MA1 AfontovaGora3 -0.0243 -2.587 144086
Mbuti Kostenki14 MA1 AfontovaGora3 -0.0211 -2.421 159767

Matt said...

But yes I personally cannot fully understand the meaning of:

Mbuti AfontovaGora3 Villabruna Vestonice -0.0199 -3.844 213665
Mbuti AfontovaGora3 Villabruna Kostenki14 -0.0281 -4.465 218311

Mbuti MA1 Villabruna Vestonice -0.0061 -1.388 531468
Mbuti MA1 Villabruna Kostenki14 -0.0124 -2.339 644492

Villabruna shows a shift towards AfontovaGora3 and MA1 relative to Vestonice, and a bigger shift towards AfontovaGora3 than MA1, implying gene flow relating to AfontovaGora3 and Villabruna (or an opposite effect in Vestonice and Kostenki14).

at the same time as:

Malta1 AfontovaGora3 Villabruna Mbuti -0.0004 -0.6 459597
Malta1 AfontovaGora3 Vestonice16 Mbuti 0.0010 1.5 398604
Malta1 AfontovaGora3 Kostenki14 Mbuti 0.0011 1.7 486631

Both Villabruna and Vestonice and Kostenki14 are all equally related to both MA1 and AfontovaGora3.

Olympus Mons said...

@Davidsi,
There is no Dna extracted from the southern Caucasus that I know of from 8M bc to 5M bc. That is more than 2 thousand years. Prior to Shulaveri there were in thte region the Koch and those were a completely different stock of people. And after those millennia one later see a “leveling event” of ashes and abandonment at 4.900 BC. That that is pivot. Because after that is all Ubaid/Uruk J2/j1.
So I don’t know where they came from. Either manage to cross over the Caucasus mountains and might very well be at Dnieper basin just the other side, that is a couple hundreds of kilometers or they even can be part of the natufian. At least they seem to share some architectural traits but I don’t go as far.
Or maybe they were a stock close to Villabruna that took routs there mixing with lots of Kotias Mtdna womem as well as middle eastern others.
To me physical anthropologically is obvious. They where the shulaveri, then at 4,900 they were in Tell tsaf, then by 4,500 bc in Merimde Eygpt, then in Iberia coalescing to bell beaker. Or not.

That is why is going to be so awesome. Time will tell.

Davidski said...

Matt,

If you want to test for pulses of admixture, you stick the potential source of the admixture next to the outgroup. Like this from the Mathieson paper, when they were testing for WHG gene flow into LBK vs the Anatolians.

Chimp WHG LBK_EN Anatolia_Neolithic -6.8

So if we want to see if Villabruna has admixture from AG3 that Vestonice lacks, or at least has a lot less of, we do the same...

Chimp AfontovaGora3 Villabruna Vestonice -5.792

Yes, it does. And we can also check if AG3 has admixture from Villabruna that MA1 lacks.

Chimp Villabruna AfontovaGora3 MA1 -0.484

Nope, it doesn't. And that's what your stats are showing as well, because they're of the same form, except presented the other way around.

Malta1 AfontovaGora3 Villabruna Mbuti -0.6
Malta1 AfontovaGora3 Vestonice16 Mbuti 1.5
Malta1 AfontovaGora3 Kostenki14 Mbuti 1.7

Matt said...

Davidski: If you want to test for pulses of admixture, you stick the potential source of the admixture next to the outgroup. Like this from the Mathieson paper, when they were testing for WHG gene flow into LBK vs the Anatolians.

I think you would still get a signal though. Take D(Mbuti, Anatolia_Neolithic, Villabruna, El Miron). That does not show to show a 0 signal of relatedness between Stuttgart and Villabruna, just because Stuttgart, which is next to the outgroup, is not the population admixing into Villabruna. D( Mbuti Anatolia_Neolithic Villabruna ElMiron) = -0.0205 -6.98 589228 per your last stats.

IIUC there should still be a non-equal signal, regardless of whether the admixeed or admixing population is with the outgroup, so long as there is some admixture between the population next to the outgroup and one of the other two, regardless.

Davidski said...

Actually, the proper way to test two populations against each other as potential mixture sources is like this...

Vestonice Villabruna AfontovaGora3 MA1 -0.0138 -2.003 178192

So assuming that Vestonice is a decent proxy for the pre-admixed Villabruna, then we almost have a significant outcome in favor of AG3. The number of markers is low, so with more markers I'd say we'd have AG3 as the more plausible source of admixture.

epoch2013 said...

@David

What would this do?

GoyetQ116-1 Vestonice16; Villabruna, Hungarian_HG

We removed GoyetQ116-1 that way as it is outgroup. Would that be an interesting way to see if Vestonice16, which shows some affinity with WHG, contibuted to WHG? Villabruna being the highest Aurignacian, lowest Asian and KO1 the opposite.

epoch2013 said...

@David

Also this: Kostenki14 as outgroup as it has a similar affinity to all UP Europeans. So this might be the way to select anything unique to Gravettian:

Kostenki14, Vestonice16; Villabruna, AfontovaGora3
Kostenki14, Vestonice16; Villabruna, El_Miron
Kostenki14, Vestonice16; Villabruna, Hungarian_HG
Kostenki14, Vestonice16; Villabruna, Karelia_HG

And any other idea that you may come up with.

Davidski said...

@Matt

I think you would still get a signal though.

I don't know why there isn't a signal there. The only thing I can think of is more of some type of western admixture in MA1 than in AG3, which is suggested by these stats.

Mbuti GoyetQ116-1 MA1 AfontovaGora3 -0.0243 -2.587 144086
Mbuti Kostenki14 MA1 AfontovaGora3 -0.0211 -2.421 159767

epoch2013 said...

Also, as K14 is almost similarly affiliated to Euro UP, would this be a way to see Gravettian admixture in WHG?

GoyetQ116-1 Vestonice16; Kostenki14, Villabruna
GoyetQ116-1 Vestonice16; El_Miron, Villabruna




epoch2013 said...

Or the other way around:

GoyetQ116-1 , Villabruna; Vestonice16, Kostenki14

ryukendo kendow said...

Time to analyse some of the stats David has kindly run.
EEF seems much closer to Villabruna than the rest, implying WHG input from Villabruna, which is confirmed in the Treemix.
Mbuti Anatolia_Neolithic Villabruna ElMiron -0.0205 -6.98 589228
Mbuti Anatolia_Neolithic Villabruna GoyetQ116-1 -0.0392 -11.459 658130
Mbuti Anatolia_Neolithic Villabruna Kostenki14 -0.0449 -13.697 905480
Mbuti Anatolia_Neolithic Villabruna Vestonice -0.0302 -12.596 736405
Small amt of Villabruna in the Vestonice and Elmiron sufficient to attract Anatolia_Neolithic compared to Kostenki or Goyet.
Mbuti Anatolia_Neolithic ElMiron GoyetQ116-1 -0.0203 -4.769 479322
Mbuti Anatolia_Neolithic GoyetQ116-1 Kostenki14 -0.0084 -1.949 631187
Mbuti Anatolia_Neolithic GoyetQ116-1 Vestonice 0.0088 2.342 574875
Mbuti Anatolia_Neolithic ElMiron Kostenki14 -0.0278 -6.878 563460
Mbuti Anatolia_Neolithic ElMiron Vestonice -0.0132 -3.989 534060
Mbuti Anatolia_Neolithic Kostenki14 Vestonice 0.0179 4.971 708579

Karelian is indeed implicated in the Villabruna phenomenon.
Mbuti Karelia_HG Villabruna ElMiron -0.0434 -9.616 568639
Mbuti Karelia_HG Villabruna GoyetQ116-1 -0.0654 -12.425 625624
Mbuti Karelia_HG Villabruna Kostenki14 -0.0733 -14.719 827456
Mbuti Karelia_HG Villabruna Vestonice -0.0619 -16.221 701000
Same behaviour as in the EEFs, small amt of Villabruna in ElMiron and Kostenice attracts Karelian.
Mbuti Karelia_HG ElMiron GoyetQ116-1 -0.0296 -4.762 466724
Mbuti Karelia_HG ElMiron Kostenki14 -0.0362 -5.817 543495
Mbuti Karelia_HG ElMiron Vestonice -0.0231 -4.644 520449
Mbuti Karelia_HG GoyetQ116-1 Kostenki14 -0.0099 -1.525 600100
Mbuti Karelia_HG GoyetQ116-1 Vestonice 0.0059 1.067 555217
Mbuti Karelia_HG Kostenki14 Vestonice 0.0153 2.848 674697

Villabruna<----->Karitiana to the exclusion of all others.
Mbuti Karitiana Villabruna ElMiron -0.0089 -2.279 589338
Mbuti Karitiana Villabruna GoyetQ116-1 -0.0127 -2.887 658456
Mbuti Karitiana Villabruna Kostenki14 -0.0132 -3.16 907963
Mbuti Karitiana Villabruna Vestonice -0.0157 -4.594 736700
Mbuti Karitiana ElMiron GoyetQ116-1 -0.0047 -0.912 479333
Mbuti Karitiana ElMiron Kostenki14 -0.0059 -1.181 563521
Mbuti Karitiana ElMiron Vestonice -0.0077 -1.743 534074
Mbuti Karitiana GoyetQ116-1 Kostenki14 -0.0004 -0.081 631360
Mbuti Karitiana GoyetQ116-1 Vestonice -0.0018 -0.369 574904
Mbuti Karitiana Kostenki14 Vestonice -0.0015 -0.35 708737

Davidski said...

epoch,

No need to write the commas, in fact better if you don't.

GoyetQ116-1 Vestonice16 VillabrunaI9030 Hungary_HG -0.0066 -0.756 505923
Kostenki14 Vestonice16 VillabrunaI9030 AfontovaGora3 -0.0309 -2.784 200873
Kostenki14 Vestonice16 VillabrunaI9030 ElMiron 0.0003 0.038 501759
Kostenki14 Vestonice16 VillabrunaI9030 Hungary_HG -0.0046 -0.548 551653
Kostenki14 Vestonice16 VillabrunaI9030 Karelia_HG -0.0226 -2.736 622303
GoyetQ116-1 Vestonice16 Kostenki14 VillabrunaI9030 0.0079 0.854 531369
GoyetQ116-1 Vestonice16 ElMiron VillabrunaI9030 0.0889 9.906 456908
GoyetQ116-1 VillabrunaI9030 Vestonice16 Kostenki14 -0.0309 -3.179 531369

ryukendo kendow said...

As for the ANE in Villabruna question, I agree with David that the X1, X2; Y1, Y2 form of stats is the best to check; I asked for this two days ago and we indeed get the signal between AG3 and Villabruna, which also matches what we see when Villabruna is closer to Karitiana.

AfontovaGora3 MA1 Villabruna ElMiron 0.0092 1.116 156191
AfontovaGora3 MA1 Villabruna GoyetQ116-1 0.0225 2.443 164195
AfontovaGora3 MA1 Villabruna Kostenki14 0.02 2.218 182581
AfontovaGora3 MA1 Villabruna Vestonice 0.0138 2.002 178192
AfontovaGora3 MA1 ElMiron GoyetQ116-1 0.0113 0.975 138333
AfontovaGora3 MA1 ElMiron Kostenki14 0.0128 1.152 145183
AfontovaGora3 MA1 ElMiron Vestonice 0.0099 1.035 149426
AfontovaGora3 MA1 GoyetQ116-1 Kostenki14 -0.0075 -0.625 153385
AfontovaGora3 MA1 GoyetQ116-1 Vestonice -0.0082 -0.822 155129
AfontovaGora3 MA1 Kostenki14 Vestonice -0.0079 -0.801 167134

However Matt's points are valid. David, your stats are nice and would contribute to resolving Matt's issue but there's a discrepancy
Mbuti Kostenki14 MA1 AfontovaGora3 -0.0211 -2.421 159767
Malta1 AfontovaGora3 Kostenki14 Mbuti 1.7

What's going on?

Chad Rohlfsen said...

I still think we're missing a ghost from West Asia that is the one dropping I-IJ. No doubt there is probably AG3 in Villabruna and beyond, but WHG is significantly closer to the UP than it is to AG3. Maybe, 20% input from AG3 and a 40/40 split of UP and a new West Asian UP

ryukendo kendow said...

@ Alberto

Alberto, KO1's WHG is closest to the WHG in EEFs is what I mean. This was the conclusion of the paper where Anatolia Neolithic was presented, but I can't remember its name.

Davidski said...

This is what I'm getting using the same dataset.

Mbuti Kostenki14 MA1 AfontovaGora3 -0.0211 -2.412 159767
MA1 AfontovaGora3 Kostenki14 Mbuti 0.0211 2.412 159767

Who ran this? If it was me, I probably used a different, smaller dataset.

Malta1 AfontovaGora3 Kostenki14 Mbuti 1.7

ryukendo kendow said...

Another thing that bothers me is that, if the contribution really is that large from AG3 to Villabruna as seen in the treemix, the Z scores should be much larger than what we see here. To be fair, all the figures are below Z>3, so the authors discounted them.

Davidski said...

Running the Villabruna cluster vs the Vestonice cluster with Chimp as outgroup I'm getting well over 3. Can't see anything wrong with this stat.

Chimp AfontovaGora3 Villabruna Vestonice -5.792

ryukendo kendow said...

David, the sites with ancestral alleles for Vestonice and Chimp due to Neanderthal in Vestonice would confound this stat and make it more negative, but to be fair, you also gave these:
AfontovaGora3 Mbuti Villabruna Kostenki14 0.0281 4.465 218311
AfontovaGora3 Mbuti Villabruna Vestonice 0.0199 3.844 213665

Also, if the branch point is so far back that its a tritomy between AG3 and Kotias ANE, then we shouldn't see such a high figure either.

The decision to use only Mal'ta1 and not AG3 in the paper during cluster modelling and in the ADMIXTUREGRAPH was probably a horrible idea, since the signals for Mal'ta are deceptively weak and led them to completely overlook this.

epoch2013 said...

Kostenki14 Vestonice16 VillabrunaI9030 ElMiron 0.0003 0.038 501759
GoyetQ116-1 Vestonice16 ElMiron VillabrunaI9030 0.0889 9.906 456908

How strange. How can that be? El Miron is only one third proto-WHG according to the papers fig. 4, yet according to Vestonice16 it is similar. However, if you substract GoyetQ116-1 from Vestonice16 it really fancies Villabruna!

Does that mean that part of WHG is that third from El Miron and *another* part - don't know how large though - comes from Vestonice?



Davidski said...

rk,

Chimp has been used in some of the major papers, so it can't be all the problematic unless someone's discovered something since the Mathieson paper. But yeah, anyway, AG3 produces significant Z scores with Mbuti as well, so the AG3-related admix into Villabruna looks legit.

Also, the D-stats and Z scores will be different when based on different numbers of markers, but unless the difference is really huge, then non-significant Zs should still be non-significant, and significant Zs should still be significant.

epoch,

I have no idea how to read those.

Davidski said...

David, the sites with ancestral alleles for Vestonice and Chimp due to Neanderthal in Vestonice would confound this stat and make it more negative.

Ah, I see now what you mean.

Chad Rohlfsen said...

On qpAdm a simple mix of Vestonice and AG3 fails miserably. F3ratios are also ridiculous failures. I'm telling you guys. European UP + West Asian migrant + AG3 = WHG.

left pops:
Villabruna
Vestonice16
AfontovaGora3

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim
Kotias
Anatolia_Neolithic

0 Villabruna 1
1 Vestonice16 1
2 AfontovaGora3 1
3 Mbuti 10
4 Dai 10
5 Biaka 12
6 Yoruba 20
7 Papuan 14
8 Ust_Ishim 1
9 Kotias 1
10 Anatolia_Neolithic 25
jackknife block size: 0.050
snps: 1151062 indivs: 96
number of blocks for block jackknife: 713
dof (jackknife): 627.428
numsnps used: 190083
codimension 1
f4info:
f4rank: 1 dof: 6 chisq: 45.239 tail: 4.19618812e-08 dofdiff: 8 chisqdiff: -45.239 taildiff: 1
B:
scale 1.000
Dai -0.743
Biaka -0.013
Yoruba -0.043
Papuan -0.962
Ust_Ishim 0.195
Kotias 0.074
Anatolia_Neolithic 2.341
A:
scale 615.709
Vestonice16 -0.539
AfontovaGora3 -1.308


full rank 1
f4info:
f4rank: 2 dof: 0 chisq: 0.000 tail: 1 dofdiff: 6 chisqdiff: 45.239 taildiff: 4.19618812e-08
B:
scale 1.000 1.000
Dai -0.177 1.455
Biaka -0.102 -0.156
Yoruba -0.020 -0.021
Papuan -0.599 1.036
Ust_Ishim 0.190 -0.515
Kotias 0.924 1.850
Anatolia_Neolithic 2.390 -0.314
A:
scale 552.392 1067.297
Vestonice16 -0.861 -1.122
AfontovaGora3 -1.122 0.861


best coefficients: 1.701 -0.701
ssres:
-0.003205204 0.000491869 0.000074793 -0.001703195 0.000978301 -0.005485799 -0.002192948
-1.207984363 0.185376544 0.028188148 -0.641903755 0.368704250 -2.067500213 -0.826483128

Jackknife mean: 1.674923243 -0.674923243
std. errors: 0.232 0.232

error covariance (* 1000000)
53904 -53904
-53904 53904


fixed pat wt dof chisq tail prob
00 0 6 45.239 0 1.701 -0.701 infeasible
01 1 7 74.194 2.09029e-13 1.000 0.000
10 1 7 225.207 0 0.000 1.000
Vestonice16 AfontovaGora3 Villabruna 535577600283907.625000 2247575462361.664551 238.291 26801
GoyetQ116-1 AfontovaGora3 Villabruna 552433054793374.125000 2245100394508.776367 246.062 27644
Kostenki14 AfontovaGora3 Villabruna 606712998430208.000000 2425659384722.072754 250.123 30360

epoch2013 said...

@ Chad

Even if you swap Vestonice for El Miron?

Davidski said...

Villabruna is probably a mix of something like AG3 and the Balkan version of Vestonice, neither of which we have yet.

But in your right pops you've got Kotias and Anatolia Neolithic. The former is in large part AG3-like and the latter has direct admixture from Villabruna. So I don't think that'll work.

Alberto said...

@RK

- Ah, ok, I misunderstood you about KO1. Then we agree about that.

- I also think there something strange with these stats:

Malta1 AfontovaGora3 Kostenki14 Mbuti 0.0011 1.7 486631
Mbuti Kostenki14 MA1 AfontovaGora3 -0.0211 -2.421 159767

The value of the one with less markers is 19x higher! (when the D value should not be too sensitive to the number of markers as the Z score).

- Re: AG3 <--> Villabruna, I have the same doubts as Matt. If the signal is there it should show up in all stats. Otherwise what it looks like is more like a drop in relatedness to pre-LGM Europeans in AG3.

- In any case, how would you distinguish AG3 admixture in Villabruna from Villabruna admixture in AG3? We'd need an unadmixed sample in the Villabruna cluster to prove AG3 admixture in the ones we have (?). That's why the stats where there is a difference between AG3 and MA1 in relatedness to Villabruna would suggest a Villabruna -> AG3 admixture (and not the other way around), but then a direct comparison with MA1 and AG3 shows no sign of gene flow between AG3 and Villabruna.

So the question wouldn't rather be: is there Villabruna admixture in AG3 or there isn't? I don't see how we could even attempt to prove AG3 admixture in Villabruna (maybe AdmixtureGraph could be of help here, since it's quite smart, using a lot of different stats, but otherwise no idea how).

ryukendo kendow said...

Epoch, you can't subtract Goyet from Vestonice. Vestonice has ancestry from Kostenki, not Goyet.
All the stats make sense when compared to the ADMIXTUREGRAPH tree, its good to familiarise yourself with it.

epoch2013 said...

@David

"But yeah, anyway, AG3 produces significant Z scores with Mbuti as well, so the AG3-related admix into Villabruna looks legit."

The significant Asian admixture in WHG in fig 4 of the paper is consistently accompanied by significant American admixture.

Davidski said...

Where is this stat from?

Malta1 AfontovaGora3 Kostenki14 Mbuti 0.0011 1.7 486631

How is it possible to get that many markers for AG3.damage? It must be more than just SNPs.

Davidski said...

Have a look at extended data table 1. It shows that only 286,355 SNPs are covered for AG3.

epoch2013 said...

@ ryukendo kendow

"Epoch, you can't subtract Goyet from Vestonice. Vestonice has ancestry from Kostenki, not Goyet."

OK. But I would suspect the score to be neutral then, as it is with K14, Vestonice.

Never mind, its probably a bogus result.

Matt said...

Davidski: Where is this stat from? - It's from p44 of the Supplementary Information. 486631 is "Sites used".

Davidski said...

They used full sequence data. No wonder the D and Z are different from those based on the "damage" SNPs. But the end result is the same; non significant Z.

Davidski said...

Chad,

Using these right pops I get Loschbour as 7.7% more AG3-like than Villabruna.

https://drive.google.com/file/d/0B9o3EYTdM8lQeWtTTGJRb0dXNzQ/view?usp=sharing

Looks like there were variable levels of North Eurasian ancestry in the Villabruna cluster.

Onur said...

@Davidski

Using these right pops I get Loschbour as 7.7% more AG3-like than Villabruna.

https://drive.google.com/file/d/0B9o3EYTdM8lQeWtTTGJRb0dXNzQ/view?usp=sharing

Looks like there were variable levels of North Eurasian ancestry in the Villabruna cluster.


So, the more ANE mix a WHG individual has, the more East Asian/Amerindian-shifted he/she is? Is the East Asian/Amerindian-shift in WHG due to the ANE mix?

Davidski said...

I don't know. Does the paper show Loschbour as more East Asian/Amerindian than Villabruna? From memory it does.

By the way, here's another one. It shows Loschbour as 21.4% more AG3-like than GoyetQ116-1.

https://drive.google.com/file/d/0B9o3EYTdM8lQVXJzRHpOM180Zjg/view?usp=sharing

Alberto said...

@Davidski

Yes, that's the way to go. If you want to find AG3 admixture in Villabruna cluster, you have to test each member to see if ones get a significant stat when compared to others.

That's how we can prove GoyetQ116 related admixture in Loschbour or LaBrana1. Because we have Villabruna and KO1 that lack such admixture (or have less). And that's how we can prove Villabruna admixture in ElMiron, because we have other samples from her cluster who lack such admixture.

So we can model Loschbour as Villabruna + GoyetQ116 (or better some member of ElMiron cluster) and we can model ElMiron as one member of her cluster + Villabruna.

So for example, can we prove or disprove that KO1 has already some GoyetQ116 admixture? No, we can't. It doesn't mean he doesn't have it, but until we find a sample that forms a clade with KO1 but with less affinity to GoyetQ116, it's futile to speculate about it.

Same goes for AG3 admixture in Villabruna. We need to invent yet another ghost population without such hypothetical admixture to be able to fit it. I can't see the point in doing so.

Davidski said...

I'm not sure what the problem is, since the Villabruna cluster is closely related to the Vestonice cluster. They're basically sister clades, but Vestonice is older. So we can test Villabruna against Vestonice, like in the Treemix, and the results show significant AfontovaGora3 affinity in Villabruna.

Moreover, it's clear that there are variable levels of this affinity within the Villabruna cluster, because the samples within this cluster show variable levels of affinity to East Asians and Amerindians. This was shown in the paper.

So Villabruna has AG3/Amerindian/East Asian related admixture, and it looks fairly recent.

Onur said...

@Davidski

I don't know. Does the paper show Loschbour as more East Asian/Amerindian than Villabruna? From memory it does.

Yes. In fact, Loschbour shows the highest East Asian/Amerindian shift among the WHG samples (see Figure 4b in the main paper).

Chad Rohlfsen said...

Alright, I'm changing some things up and running a few more. I'll post them in a few minutes.

Chad Rohlfsen said...

left pops:
Villabruna
Vestonice16
AfontovaGora3

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim

0 Villabruna 1
1 Vestonice16 1
2 AfontovaGora3 1
3 Mbuti 10
4 Dai 10
5 Biaka 12
6 Yoruba 20
7 Papuan 14
8 Ust_Ishim 1
jackknife block size: 0.050
snps: 1150832 indivs: 70
number of blocks for block jackknife: 713
dof (jackknife): 627.426
numsnps used: 190142
codimension 1
f4info:
f4rank: 1 dof: 4 chisq: 5.411 tail: 0.247697934 dofdiff: 6 chisqdiff: -5.411 taildiff: 1
B:
scale 1.000
Dai 1.666
Biaka -0.135
Yoruba -0.017
Papuan 1.352
Ust_Ishim -0.614
A:
scale 1176.550
Vestonice16 -0.776
AfontovaGora3 1.182


full rank 1
f4info:
f4rank: 2 dof: 0 chisq: 0.000 tail: 1 dofdiff: 4 chisqdiff: 5.411 taildiff: 0.247697934
B:
scale 1.000 1.000
Dai 1.475 -1.522
Biaka -0.025 0.811
Yoruba 0.004 0.160
Papuan 1.548 1.409
Ust_Ishim -0.654 -0.127
A:
scale 1108.344 3423.918
Vestonice16 -0.500 1.323
AfontovaGora3 1.323 0.500


best coefficients: 0.604 0.396
ssres:
-0.000147515 0.000230989 0.000047413 0.000720389 -0.000168044
-0.417380567 0.653563232 0.134151194 2.038279717 -0.475465894

Jackknife mean: 0.600101915 0.399898085
std. errors: 0.136 0.136

error covariance (* 1000000)
18548 -18548
-18548 18548


fixed pat wt dof chisq tail prob
00 0 4 5.411 0.247698 0.604 0.396
01 1 5 12.419 0.0294748 1.000 0.000
10 1 5 22.159 0.000488453 0.000 1.000
best pat: 00 0.247698 - -
best pat: 01 0.0294748 chi(nested): 7.009 p-value for nested model: 0.00811234

Chad Rohlfsen said...

Karelia_HG failed as this mixture, so I'll need to change things up. I'll throw some more out here as I go.

Chad Rohlfsen said...

left pops:
Villabruna
GoyetQ116-1
Vestonice16
AfontovaGora3

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim

0 Villabruna 1
1 GoyetQ116-1 1
2 Vestonice16 1
3 AfontovaGora3 1
4 Mbuti 10
5 Dai 10
6 Biaka 12
7 Yoruba 20
8 Papuan 14
9 Ust_Ishim 1
jackknife block size: 0.050
snps: 1150853 indivs: 71
number of blocks for block jackknife: 713
dof (jackknife): 626.322
numsnps used: 168323
codimension 1
f4info:
f4rank: 2 dof: 3 chisq: 4.935 tail: 0.176646786 dofdiff: 5 chisqdiff: -4.935 taildiff: 1
B:
scale 1.000 1.000
Dai 1.651 -0.277
Biaka -0.150 0.121
Yoruba 0.016 0.145
Papuan 1.467 0.076
Ust_Ishim 0.313 2.210
A:
scale 1264.163 1816.350
GoyetQ116-1 0.868 1.466
Vestonice16 -0.852 0.359
AfontovaGora3 1.234 -0.850


full rank 1
f4info:
f4rank: 3 dof: 0 chisq: 0.000 tail: 1 dofdiff: 3 chisqdiff: 4.935 taildiff: 0.176646786
B:
scale 1.000 1.000 1.000
Dai 1.442 -0.441 -1.516
Biaka -0.034 0.166 0.808
Yoruba -0.027 0.083 -0.174
Papuan 1.696 0.107 1.355
Ust_Ishim 0.212 2.182 -0.427
A:
scale 1199.867 1832.113 4424.931
GoyetQ116-1 0.954 1.442 0.104
Vestonice16 -0.570 0.262 1.614
AfontovaGora3 1.328 -0.924 0.619


best coefficients: 0.064 0.580 0.356
ssres:
-0.000135128 0.000199062 -0.000054305 0.000637496 -0.000176957
-0.427933411 0.630405633 -0.171977317 2.018870155 -0.560400600

Jackknife mean: 0.101271159 0.551055068 0.347673773
std. errors: 0.458 0.287 0.236

error covariance (* 1000000)
210034 -118446 -91589
-118446 82413 36033
-91589 36033 55556


fixed pat wt dof chisq tail prob
000 0 3 4.935 0.176647 0.064 0.580 0.356
001 1 4 8.085 0.0885098 0.627 0.373 -0.000
010 1 4 9.238 0.055424 0.973 -0.000 0.027
100 1 4 4.972 0.290149 -0.000 0.617 0.383
011 2 5 9.249 0.0995483 1.000 -0.000 0.000
101 2 5 12.224 0.031849 0.000 1.000 0.000
110 2 5 23.496 0.000271268 0.000 -0.000 1.000
best pat: 000 0.176647 - -
best pat: 100 0.290149 chi(nested): 0.038 p-value for nested model: 0.84607
best pat: 011 0.0995483 chi(nested): 4.276 p-value for nested model: 0.0386469

Chad Rohlfsen said...

Loschbour may just be less Neandertal than Villabruna, causing it to be closer to ENA and ANE. David, can you check with that set you have with Neandertal?

Chad Rohlfsen said...

left pops:
ElMiron
GoyetQ116-1
AfontovaGora3

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim

0 ElMiron 1
1 GoyetQ116-1 1
2 AfontovaGora3 1
3 Mbuti 10
4 Dai 10
5 Biaka 12
6 Yoruba 20
7 Papuan 14
8 Ust_Ishim 1
jackknife block size: 0.050
snps: 1150793 indivs: 70
number of blocks for block jackknife: 713
dof (jackknife): 625.744
numsnps used: 158652
codimension 1
f4info:
f4rank: 1 dof: 4 chisq: 4.834 tail: 0.304790422 dofdiff: 6 chisqdiff: -4.834 taildiff: 1
B:
scale 1.000
Dai -0.728
Biaka 0.023
Yoruba -0.013
Papuan -0.619
Ust_Ishim 2.021
A:
scale 1513.965
GoyetQ116-1 0.764
AfontovaGora3 -1.190


full rank 1
f4info:
f4rank: 2 dof: 0 chisq: 0.000 tail: 1 dofdiff: 4 chisqdiff: 4.834 taildiff: 0.304790422
B:
scale 1.000 1.000
Dai 1.309 -0.398
Biaka 0.185 0.066
Yoruba 0.535 0.134
Papuan 1.693 -0.148
Ust_Ishim 0.315 2.190
A:
scale 1473.763 1545.386
GoyetQ116-1 0.747 1.201
AfontovaGora3 1.201 -0.747


best coefficients: 0.609 0.391
ssres:
0.000708381 0.000134903 0.000373952 0.001019854 0.000819352
1.028632905 0.195890922 0.543011775 1.480918958 1.189772392

Jackknife mean: 0.589453591 0.410546409
std. errors: 0.258 0.258

error covariance (* 1000000)
66359 -66359
-66359 66359


fixed pat wt dof chisq tail prob
00 0 4 4.834 0.30479 0.609 0.391
01 1 5 7.622 0.178337 1.000 0.000
10 1 5 11.282 0.0460678 0.000 1.000
best pat: 00 0.30479 - -
best pat: 01 0.178337 chi(nested): 2.788 p-value for nested model: 0.0949561

Chad Rohlfsen said...

left pops:
ElMiron
GoyetQ116-1
Vestonice16
AfontovaGora3

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim

0 ElMiron 1
1 GoyetQ116-1 1
2 Vestonice16 1
3 AfontovaGora3 1
4 Mbuti 10
5 Dai 10
6 Biaka 12
7 Yoruba 20
8 Papuan 14
9 Ust_Ishim 1
jackknife block size: 0.050
snps: 1150829 indivs: 71
number of blocks for block jackknife: 713
dof (jackknife): 624.282
numsnps used: 146272
codimension 1
f4info:
f4rank: 2 dof: 3 chisq: 4.111 tail: 0.24976898 dofdiff: 5 chisqdiff: -4.111 taildiff: 1
B:
scale 1.000 1.000
Dai 1.762 -0.069
Biaka -0.305 0.010
Yoruba -0.116 -0.007
Papuan 1.329 0.034
Ust_Ishim -0.149 2.235
A:
scale 1353.237 1883.614
GoyetQ116-1 0.028 1.293
Vestonice16 -1.635 -0.434
AfontovaGora3 0.570 -1.068


full rank 1
f4info:
f4rank: 3 dof: 0 chisq: 0.000 tail: 1 dofdiff: 3 chisqdiff: 4.111 taildiff: 0.24976898
B:
scale 1.000 1.000 1.000
Dai 1.711 -0.119 -0.830
Biaka -0.225 0.135 0.976
Yoruba -0.019 0.207 1.372
Papuan 1.418 0.306 1.149
Ust_Ishim -0.089 2.198 -0.394
A:
scale 1366.163 1834.630 3512.569
GoyetQ116-1 0.399 1.608 0.506
Vestonice16 -1.284 -0.046 1.161
AfontovaGora3 1.091 -0.643 1.181


best coefficients: 0.410 0.158 0.432
ssres:
0.000305372 0.000206526 0.000388365 0.000806926 0.000319040
0.669685956 0.452914280 0.851690080 1.769601666 0.699659437

Jackknife mean: 0.392531507 0.168271465 0.439197028
std. errors: 0.436 0.247 0.260

error covariance (* 1000000)
189841 -91833 -98007
-91833 61250 30584
-98007 30584 67423


fixed pat wt dof chisq tail prob
000 0 3 4.111 0.249769 0.410 0.158 0.432
001 1 4 7.421 0 1.061 -0.061 -0.000 infeasible
010 1 4 4.689 0.320711 0.611 -0.000 0.389
100 1 4 5.612 0.230052 0.000 0.361 0.639
011 2 5 7.473 0.18777 1.000 -0.000 0.000
101 2 5 24.014 0.000215798 0.000 1.000 0.000
110 2 5 11.150 0.0484843 0.000 -0.000 1.000
best pat: 000 0.249769 - -
best pat: 010 0.320711 chi(nested): 0.579 p-value for nested model: 0.446894
best pat: 011 0.18777 chi(nested): 2.784 p-value for nested model: 0.095214

Chad Rohlfsen said...

left pops:
ElMiron
GoyetQ116-1
Villabruna

right pops:
Mbuti
Dai
Biaka
Yoruba
Papuan
Ust_Ishim

0 ElMiron 1
1 GoyetQ116-1 1
2 Villabruna 1
3 Mbuti 10
4 Dai 10
5 Biaka 12
6 Yoruba 20
7 Papuan 14
8 Ust_Ishim 1
jackknife block size: 0.050
snps: 1150844 indivs: 70
number of blocks for block jackknife: 713
dof (jackknife): 632.607
numsnps used: 433657
codimension 1
f4info:
f4rank: 1 dof: 4 chisq: 5.598 tail: 0.231233451 dofdiff: 6 chisqdiff: -5.598 taildiff: 1
B:
scale 1.000
Dai 0.966
Biaka -0.303
Yoruba -0.260
Papuan 1.346
Ust_Ishim 1.447
A:
scale 2452.359
GoyetQ116-1 0.572
Villabruna -1.293


full rank 1
f4info:
f4rank: 2 dof: 0 chisq: 0.000 tail: 1 dofdiff: 4 chisqdiff: 5.598 taildiff: 0.231233451
B:
scale 1.000 1.000
Dai 0.561 -1.109
Biaka 0.027 0.792
Yoruba 0.261 1.278
Papuan 0.974 -1.033
Ust_Ishim 1.915 0.665
A:
scale 1898.958 4123.456
GoyetQ116-1 1.376 0.328
Villabruna -0.328 1.376


best coefficients: 0.693 0.307
ssres:
0.000077814 0.000136967 0.000318667 0.000275235 0.000965200
0.163413121 0.287638115 0.669217784 0.578007631 2.026971929

Jackknife mean: 0.533257793 0.466742207
std. errors: 0.751 0.751

error covariance (* 1000000)
564485 -564485
-564485 564485


fixed pat wt dof chisq tail prob
00 0 4 5.598 0.231233 0.693 0.307
01 1 5 6.016 0.304666 1.000 -0.000
10 1 5 7.909 0.16135 0.000 1.000
best pat: 00 0.231233 - -
best pat: 01 0.304666 chi(nested): 0.418 p-value for nested model: 0.518031

Chad Rohlfsen said...

I'm not sure about this and the results. This makes Magdalenian 18kya more AG3 like than Villabruna 14kya.

Chad Rohlfsen said...

There has to be a ghost pop, or something. We need Balkan and West Asian UP and epi-Paleolithic samples.

Rokus said...

Villabruna is a sister clade of the earlier European Vestonice clade
The paper explains that Villabruna is genetically more related to GoyetQ116-1. I guess you meant to refer to some eastern admixtures to account for Villabruna R1b?

Villabruna and/or closely related foragers contributed significant ancestry to Neolithic Anatolians
Before this feature was interpreted as an omnipresent WHG component.

Davidski said...

Matt,

I added MA1 to the trees. Check out the update above.

Very interesting results.

Rob said...

@ Epoch (& anyone interested about Siberia <-> Europe in LGM)

Here is an interesting illustration


It had been speculated there were Giant inland lakes, but the project appears to have never taken off.

@ Chad

"We need Balkan and West Asian UP and epi-Paleolithic samples. "

Yep - & Russia-Ukraine. I think that's the only way to clarify everything.
Hopefully we don't have to wait 5 years for this ?

a said...

The unstoppable Italian machine might have a hole in the sack.
The Eastern clans set forth challenge.What to do? The suspense.
Georgian sample looks interesting in light of the Villabruna run with a pinch of ancient Kotias cluster.
R1b-M343+, P25+, P297-, V88-, M269-, M18-, M335-, M73-

Davidski said...

Here's another tree that throws up more details, including Neanderthal admixture in the older European samples.

https://drive.google.com/file/d/0B9o3EYTdM8lQWVFPdElzR2g0Wmc/view?usp=sharing

Villabruna to me looks like a composite of Vestonice, Balkan foragers, and post-Kostenki Eastern European foragers, maybe with some minor Caucasus forager input.

Complex stuff, but it actually makes sense.

Rob said...

Dave what does Villabruna look like if you use the Italian Gravettian sample instead of the Vestonice one (if permissible by quality) ?

Davidski said...

Not enough markers.

Rob said...

Shame. I'd imagine the Italian Gravettians might give a slightly better fit what we're looking for into Villabruna, but probably won't change anything overall.

Rokus said...

Anyway, all attempts to twist Villabruna towards an eastern origin closer to Vestonice, despite the Fu2016 paper, don't seem to take into consideration that 14 kya Villabruna R1b closely coincides with the very age of R1b. Maybe by then it didn't even exist anywhere else but here.

Davidski said...

I see. And then it migrated to the Eastern European steppe. So Samara HG is actually from Italy.

Thanks. Makes perfect sense.

Gioiello said...

The poem of the great Italian poet Attilio Bertolucci, father of the more famous son, the cineaste Bernardo Bertolucci, "La capanna indiana", ends with these verses: "Qui siamo giunti dove volevamo" (We came here where we wanted").
Bertolucci was born in Emilia, but his family, as also his surname says, came from Tuscany ("Dalle Maremme coi cavalli/ From Maremma with horses").

epoch2013 said...

@Davidski

Not "migrated" but more like exchanged. It's not impossible, nor is it more improbable than ANE migrating westward. Does WHG prefer Karelia over Samare HG or vice versa? Have we tried that before? If not:

Mbuti Villabruna Karelia_HG Samara_HG
Mbuti Hungarian_HG Karelia_HG Samara_HG

@Rob

Ik keep thinking this: Migration for a HG means following the herds. According to a paper Magdalenians hunted horses, and according to their art also reindeer. The latter would have keep going north when the ice retreated. So it is logical that HGs from the Magdalenian went north. We know their culture: Ahrensberg Culture.

Gravettian art is made of ivory and they hunted mammoths. That would mean there is a rationale for Vestonice to come up with Kostenki14 ancestry: Both were mammoth hunters.

http://www.archaeology.org/news/2729-141124-mammoth-consumption-hunting

Mesolithic HG's were known for their switch to small game and fish. Is there any conclusion we can draw from that? What would be their incentive for spreading over Europe? Or is this an unnecessary consideration?

epoch2013 said...

@Rob

The Magdalenian paper:

http://www.sciencedirect.com/science/article/pii/S1040618214000299

Davidski said...

epoch,

Mbuti Villabruna Karelia_HG Samara_HG 0.0064 0.929 364859
Mbuti Hungary_HG Karelia_HG Samara_HG 0.0024 0.348 323905

Grey said...

Thinking in terms of ecozones the mammoth steppe spread all the way to Britain via the north European plain so could Villabruna reflect a mix of two latitudinally separated populations: WHG in the southern zone and mammoth steppe in the north?

(with the steppe part bringing the R1b and the east asian component)

epoch2013 said...

EHG is now considered a hybrid of WHG and ANE. If that is true there must have been a place where the contact took place. GoyetQ116-1 is not equally distributed over WHG. Is there a way to use that?

Villabruna and Bichon are 14.000 year old, have no no less AG3, Asian, American. There were still mammoths 14.000 years ago, they died out 12.000 years ago. If ANE contacted WHG they might have contacted Gravettians before. Is there a Gravettian signal in EHG?

Mbuti Vestonice16 Hungary_HG Karelia_HG
Mbuti Vestonice16 Hungary_HG Samara_HG

Grey said...

sort of

SSSSSSSSSSSSS
WWWWWWWCCCCCC
WWWWWWWWWWWWW

where
S = mammoth steppe zone
C = CHG zone
W = WHG zone

Davidski said...

epoch,

Mbuti Vestonice16 Hungary_HG Karelia_HG -0.0324 -5.03 514560
Mbuti Vestonice16 Hungary_HG Samara_HG -0.0488 -6.534 281279

epoch2013 said...

@David

I reckon that is ANE in EHG. So, nothing discernable.

epoch2013 said...

Mind you. Thanks for the effort.

Shaikorth said...

"Villabruna and Bichon are 14.000 year old, have no no less AG3, Asian, American."

Wasn't Bichon supposed to be different from Villabruna and similar to Loschbour?

Mbuti Han Bichon Loschbour
Mbuti Han Villabruna Bichon

Mbuti Karitiana Bichon Loschbour
Mbuti Karitiana Villabruna Bichon

Mbuti AG3 Bichon Loschbour
Mbuti AG3 Villabruna Bichon

Davidski said...

Mbuti Han Bichon Loschbour 0.0035 0.883 959802
Mbuti Han Villabruna Bichon 0.008 1.875 763169

Mbuti Karitiana Bichon Loschbour 0.008 1.697 959802
Mbuti Karitiana Villabruna Bichon 0.0053 1.025 763169

Mbuti AfontovaGora3 Bichon Loschbour -0.0045 -0.639 229661
Mbuti AfontovaGora3 Villabruna Bichon 0.0009 0.109 218941

Chad Rohlfsen said...

Rokus,

The Villabruna is pre-p297, at the timeline for m269, and maybe 3k years behind where it should be in mutations for TMRCA.

Just wait until we have Balkan and Anatolian hunters from 25kya. We will see a mix of an unseen branch, AG3 and UP mixture in WHG. Villabruna shows about the same Neandertal input as UP samples. It's significant relationship to AG3 can't be explained away just yet. Just to clarify, I don't think the AG3 input is over 20%.

Shaikorth said...

Hmm, so AG3 doesn't seem to increase with time in Villabruna cluster.

Mbuti Han Villabruna Loschbour
Mbuti Karitiana Villabruna Loschbour
Mbuti AG3 Villabruna Loschbour

Gioiello said...

That I wrote to my friend Dartraighe on www.eng.molgen.net:

The R* found [Mal'ta boy] is a dead haplotype, old, thus with many no calls, and doesn't demonstrate anything. As the sample of Villabruna (some stupid person has written Vollabrun on Anthrogenica: they are all coming out mad with their being against Italians: Villabruna is an Italian name and its origin may be linked to the Rhaetians-Etruscans: is there anyone who remembers my "Rhaetian-Etruscan Fatherland?) is an intermediate sample between L278 and P297, thus very likely not our direct ancestor, but this demonstrates that there was a population of linked persons from where only one survived: the P297* from whom all the M73 and the M269 descend.

Gioiello said...

www.eng.molgen.org

Davidski said...

Mbuti Han Villabruna Loschbour 0.0107 2.673 756274
Mbuti Karitiana Villabruna Loschbour 0.0131 2.588 756274
Mbuti AfontovaGora3 Villabruna Loschbour -0.0042 -0.569 216528

Gioiello said...

Of course we cannot exclude that the R-P297* formed elsewhere from Italy. We need other aDNA from many other places, but, for responding /a/, my theory has shown that Italy has all the pathway of the R1b1 haplogroup (and of many others). Look at it is ridiculous the R1b basal subclades project which threw out an R-M335 sample from Italy I individuated from 2011. I ask myself how a good researcher who did that tree may resist there...

epoch2013 said...

@Chad

"Loschbour may just be less Neandertal than Villabruna, causing it to be closer to ENA and ANE. David, can you check with that set you have with Neandertal?"

I think you are right and you can see that in the paper. See supplementary info table: S3.1. El Miron has slightly less than Vestonci16 and Kostenki14.

huijbregts said...

@ All who have used the unscale function 'scaled2raw' in nMonte.

This function is meant to undo the effect of scaling (dividing the variables by the root of their variance).
Alas, it was too good to be true. I have to confess that I made an error with this function; I have removed it from nMonte.
This has no consequences for nMontes main function 'getMonte' which can be used to estimate the composition of the DNA as a mixture of related DNA's..

If I correctly understand the mathematics, the situation is even worse: every attempt to unscale the PCA-scores compromises the orthogonality of the principal components.
As a consequence it seems impossible to calculate a real absolute Euclidean distance between two DNA-samples.
What remains is the possibility to calculate the Euclidean distance of scaled data (the correlation matrix instead of the covariance matrix).
In doing so the weight of higher (=smaller) dimensions becomes inflated.
This is one more reason to be sparing with dimensions. (another one is that PCA's are sensitive to outliers.)

Jaydeepsinh Rathod said...

Looking at Treemix I have a suggestion. Could South Central Asia have been the source population from where migrations took place in two directions - one Northward toward Siberia resulting in ANE rich Malta boy & Afontova Gora sample, and the other westward toward the Caucasus and perhaps North Africa, eventually reaching Italy ?

Unfortunately, the paper that has just come out, as wonderful as it is, did not include South Asians for comparison with the ancient samples, while they included Near East & East Asians. The inclusion of South Asians could probably have, in my humble opinion, given a more clear picture regarding the connections between CHG, ANE-rich Malta group & the Villabruna group.

1. Among modern Eurasians, South Central Asian people have the highest amount of ANE. South Asia also has early subclades of ydna Q & has its sister clade R. Therefore it is a region where the split between Q & R might have taken place and could well be the place of origin of ANE.

2. South Central Asia is also the region where we find both the subclades of R, R1 & R2 well spread out across the subcontinent & into Central Asia in substantial numbers. Therefore, the split between R1 & R2 could also have happened in South Central Asia. The split between R1 & R2 is dated to between 32 to 28 kya. The split between R1a & R1b happened not much later between 27 to 23 kya (source - Poznik et al). Plus we have in Bhutan an R1b sample which is so basal that it is estimated to have split up immediately after the split up between R1a & R1b. In such a circumstance, the division of R1 into R1a & R1b within South Central Asia looks like a legitimate possibility. It would look slightly complicated to suggest that the split of R1 & R2 happend in SC Asia, then R1 moved out soon after, and then split up into R1a & R1b not much later and then both of them moved into SC Asia again - especially when we already have evidence of such a basal clade of R1b in Bhutan.

If R1b had migrated into Italy, among the Villabruna people, from SC Asia it would explain why Afontova Gora is close to Villabruna, because both AG & SC Asians are rich in ANE. A migration of R1b all the way from SC Asia will also explain why Villabruna lacks basal Eurasian.

3. However, the close relationship between CHG & Villabruna on one hand & CHG & MA-1 on the other also needs explaining. According to the paper, the West Eurasian component in Satsurblia forms a clade with MA-1 to the exclusion of all the ancient European samples.

This suggests an ancient link between ANE & CHG type ancestry. Besides ANE, SC Asia is also rich in CHG ancestry. At present we can only link HG J with CHG. A recent study on Indian J lineages, suggested that the presence of J2 in South Asia is quite old and widespread and is most likely dating to the pre-Neolithic era. Therefore, what this suggests is that while HG J may not have formed or separated in South Asia it has a very ancient presence in South Asia. Secondly, the Metspalu 2011 paper on South Asian populations, clearly said that the ANI like component that peaks in Caucasus & in SC Asia (Caucasus-Gedrosian) had probably already separated into the Caucasus & Gedrosian components by 12500 BP. Hence a Satsurblia like people spreading CHG in SC Asia is highly unlikely. More so due to lack of Basal Eurasian in SC Asia. Before 12500 BP, either the CHG came from the Caucasus into S Asia or it went from S Asia to Caucasus. A migration of CHG from S Asia to Caucasus along with J lineages will however neatly explain the results of the current paper as well as the Treemix.

If the West Eurasian like ancestry in Satsurblia came from SC Asia, it will explain why it forms a clade with MA-1 to the exclusion of other ancient Europeans.

epoch2013 said...

"Plus we have in Bhutan an R1b sample which is so basal that it is estimated to have split up immediately after the split up between R1a & R1b."

IIRC we have similar R-M343 in and around Iran: http://www.anthrogenica.com/showthread.php?1005-M343*-has-an-elevated-presence-in-northern-Iran-among-several-Iranian-speaking-groups

Chad Rohlfsen said...

M343 is in West Asia, Cenral Asia, Western China, and I believe a Nepali, in addition to the Bhutanese individual.

Alberto said...

And how about EHG being modeled now with AfontovaGora3? It should work better than with MA1. Would someone try qpAdm with left pops:

Samara_HG
AfontovaGora3
Loschbour
Kotias

a said...

@Gioiello said...

I'm waiting for new tree with Herc2 positive WHG.-Villabruna placement at base. :)
Let us see.

Davidski said...

Alberto,

https://drive.google.com/file/d/0B9o3EYTdM8lQVGt6bFA2YmhNSlk/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQbjlvV0dGSklXTU0/view?usp=sharing

Jaydeepsinh Rathod,

You have to keep in mind that Villabruna, AfontovaGora3, Mal'ta1 and the non-basal part of Kotias/Satsurblia are closely related to each other and to Vestonice, ElMiron and even Kostenki.

They're basically a super West Eurasian Hunter-Gatherer clade much more closely related to each other than to East Asians and the non-West Eurasian part of South Asians.

So if you posit that one or a couple of them originated in South Central Asia, then you're basically implying that they all did.

This is possible in theory, but if this migration happened, then it took place before the LGM peak (aka Ice Age), and going that far back is always very tricky. Moreover, R1a and R1b didn't exist back then, so the presence of basal R1b in and around South Asia would still look like a back migration.

I suppose you could argue that South Asia was like a reservoir of this West Eurasian super clade, and its different offshoots moved north into Europe, the Near East and Siberia every few thousand years. But as we get more samples from Europe and northern Asia, the gaps fill, and it's already pretty clear that there's no need to look to South Asia to find any relatively recent ancestral groups, because they're all there in Europe and nearby going back to the LGM at least.

Ust'-Ishim is actually very East and South Asian-like. But he's from Western Siberia and dated to 45,000 years ago. Maybe his ancestors moved through South Central Asia? But if so, this can't be linked to the emergence of West Eurasians.

Gioiello said...

To Jaydeepsinh Rathod I would say that the R1b present in India for old times are R1b1-L278+, but L389-. I was the first, when I wrote on Anthrogenica yet, to make Raza and Joshi test for this SNP L389, because , from their haplotype, I hypothesized that they were old and not linked to the Western European R1b subclades. Beyond that I said that those haplotypes came very likley from Central Asia and weren't original from India.
About R1b1-L389+ I demonstrated that the highest variamce is in Italy, and also the Caucasus has one R1b1-L389+ haplotype, but with YCAII=23-23, whereas in Italy we have 18-22, 18-23 and 23-23 (Invisible Sun from Sicily I made him test he too).
India has other R1b haplotypes, but are those downstream the R-L23 from Samara, as R-L277 (Italian Mattoli belongs to this haplotypes but he is L277 negative, thus very likely the oldest haplotypes of this subclades are in Italy they too) and some rare haplotype which may have come to India with the massive R1a from the Russian plain.

Seinundzeit said...

Jaydeepsinh Rathod,

Those are some pretty interesting ideas. I think it'll be very exciting to see what West Asian/South Central Asian aDNA has to reveal.

Regardless, at the moment, I think David's response pretty much encapsulates why these ideas seem unparsimonious (again, at the moment, as upcoming aDNA work could easily prove you right).

I think the reason South Central Asians have the highest amounts of ANE admixture out of all predominantly West Eurasian populations, is primarily due to them deriving their genetic heritage from Andronovo-like steppe populations (which had a lot of ANE, more than even the most ANE-admixed northern/eastern Europeans), CHG-like West Asian agriculturalists (who seem to have been predominately ANE, with Basal Eurasian admixture, and probably some Villabruna-related admixture), and South Asian hunter-gatherers (who seem to have had, based on evidence from contemporary populations, some kind of component related to ANE, perhaps a very basal branch in the North Eurasian group). So in theory, all three "ancestral populations" (to use cookie cutter, overly sloppy/simplistic language, for the sake of brevity) for South Central Asians had very substantial amounts of ANE ancestry.

Also, I'd say that the Mansi are the most ANE-admixed population in Eurasia. For example, a comparison between them and the Kalash/Shugnan Pamiri Tajiks, using the d-stats nMonte sheet:

Kalash
44.15% ANE
27.45% Basal Eurasian
14.90% Villabruna-related
13.50% ENA

Tajik (Shugnan)
45.10% ANE
26.05% Basal Eurasian
20.40% Villabruna-related
8.45% ENA

Mansi
56.95% ANE
33.10% ENA
9.95% Villabruna-related

The Kalash/Pamiri Tajiks do have more ANE-related admixture than any European/West Asian population, but the Mansi definitely have the most ANE-related admixture out of all populations in Eurasia.

For whatever it's worth, the most ANE-admixed living European population are the Saami, they exceed all other living European populations in terms of this sort of ancestry:

Saami
39.95% ANE
35.40% Villabruna-related
18.10% ENA
6.55% Basal Eurasian

Side note: In these fits, ANE is MA1, Villabruna-related is WHG, and "Basal Eurasian" is just a ghost I constructed using the Anatolia Neolithic and WHG d-stats. So far, my Basal Eurasian population has worked great, and is totally identical to the Lazaridis et al. construct (I've improved it further, after my first attempt).

Here are some very basic/simple fits for the Kalash/Pamiri that involve admixed populations:

Kalash
41.2% Andronovo
31.6% Caucasus_HG
27.2% Kharia

Tajik (Shugnan)
62.65% Andronovo
18.85% Kharia
18.50% Caucasus_HG

Ryan said...

"Present-day East Asians might be ancient hybrids with admixture from the same or very similar North Eurasian population, although as per the above mentioned quirks of Treemix, it's possible that the North Eurasians that contributed ancestry to Villabruna, Caucasus foragers and Eurasian steppe populations were in fact partly East Asian."

Dene-Caucasian is looking better and better isn't it.

ryukendo kendow said...

@ Davidski

David, for the most recent tree, the topology is probably not yet stable as no basal Eurasian edges appear. Could you increase the number to 8 migration edges?

ryukendo kendow said...

@ David

On another note, I think it was Wong et al who tied down migration edges by running large numbers of Treemixes and tabulating the robustness of admixture edges by their frequency. How long does each Treemix take? If we can perform Treemix with panels from which we add and drop pops in different combinations, I would be willing to help you process and tabulate some of the data if you can provide the output, but only if it doesn't fry your computer or something haha.

Alberto said...

@Davidski

Thanks! I thought that Samara_HG might get a bit of Kotias, but 24.3% Kotias? That looks a bit too high to have gone undetected so far?

Strange that for Karelia the test fails completely. No idea what sample we could be missing there.

Alberto said...

Ah, probably Motala_HG would work better than Loschbour. That was the best fit when using nMonte (but with MA1, so who knows).

Ryan said...

I'd point out that the oldest sample we have of R1b2 is from Spain, and the oldest of R1b1 is now Italy. I think you may have to revise your post about Eastern Europe as the bifurcation point of R1. I'm sure the point was somewhere on the Eurasian steppe, but there's no way there's anything directly tying it specifically to Eastern Europe anymore.

Onur said...

@Ryan

I'd point out that the oldest sample we have of R1b2 is from Spain

Do you mean the R1b1a2-V88 from Early Neolithic Spain?

Davidski said...

rk,

You mean this Treemix run?

https://drive.google.com/file/d/0B9o3EYTdM8lQdUt3N1drT2JWYW8/view?usp=sharing

Treemix doesn't take long to run, as long as the ancient samples are single genomes (as opposed to groups of individuals). In other words, if you pick a set that includes relatively high coverage ancient individuals, it won't take long to put together and run.

The other important thing to keep in mind is not get carried away with the complexity of the analysis and number of migration edges. If what you're looking for doesn't show up in the first 5 edges, then either it's not really there or you need a new tree.

If you put together a list of samples I can have a crack at it. Then I can send you the output for the number crunching.

Ryan,

Eastern Europe has Hunter-Gatherers that belong to both R1a and R1b.

Think about it.

Ryan said...

"Eastern Europe has Hunter-Gatherers that belong to both R1a and R1b."

It also has hunter-gatherers that belong to J2, but I don't think anyone here is suggesting that haplogroup IJK broke up in Eastern Europe. England contains both R1b and R1a today, but that doesn't mean England was the site of this bifurcation anymore than hunter gatherers in Karelia were.

More importantly though, Villabruna is 6,000 years older than those samples. For all we know, Karelian R1b could have come from Villabruna. We just don't have enough samples to know the route each of these haplogroups took.

Correct me if I'm wrong here, but R1b2 has not been found in Eastern Europe yet, has it?

Davidski said...

So you really believe that England and/or Italy had indigenous foragers that carried both R1a and R1b?

FrankN said...

Dave - could you run a treemix that includes

(a) Paniya - to maybe get a hint about those strange 2-3% ASI/Paniya that tend tended to show up in EHG in the nMonte runs,

(b) one or two of those East Asian (Han)-admixed WHGs - I propose here Bichon as the oldest of them, and KO1 as the least EL_Miron admixed. Here, we might gain some insight about we are actually talking one or separate East Asian incursions (KO1 fits timewise with early European millet planting, possibly a Chinese import, but Bichon would be too early for this)?

If these become too many edges, Biaka, possibly also Dai may be taken out.

Otherwise, i find the 6-8 edge runs highly interesting:

1. They point at something like a "Basal West Eurasian" split, from which not only MA1/AG3, and Kostenki/Vestonice/Vilabruna emerge, but that also contributes nearly 50% of UI and East Asian ancestry. Tentatively, this could be OOA 2, ca 50kya, which probably took the route via the Sinai. The older branchings, East Asian and CHG/AnatNeol, would then relate to the first OOA, ca 100kya, via the Arabian coast and the Persian Gulf.

2. We get two tiny admixes, namely
(a) 3% from Mbuti (directly, not the path leading to them!) into the CHG/AnatNeol bifurcation. May that relate to the spreading of the domestic dog (Basenji) from Eurasia via Israel into Africa? Dating that event is difficult, but since the Basenji is singing, not barking, his ancestors had obviously not yet been selected as (barking) herding dogs, so that transfer should pre-date Near Eastern Pre-Pottery Neolithic.
(b) 1% from Proto-CHG into Vilabruna. This may relate to the 2% "basal" signal into Villabruna picked up in the Fu study.

While those signals aren't particularly clear, I nevertheless think observing them is worthwhile, so migration edges >5 don't seem to be completely useless to me.

Ryan said...

@Davidski - "So you really believe that England and/or Italy had indigenous foragers that carried both R1a and R1b?"

You miss my point. I point to England because R1b be are both present but likely *not* indigenous. Your assumption that R1b and R1a are indigenous to Karelia and Samara is just that - an assumption. We don't know how or when each lineage arrived there, and whether they came together, separately, or simply as R1*. Agriculturalist or hunter-gatherer, the assumption of stasis and deep continuity seems to be a poor one. Perhaps Eastern Europe is the exception. Perhaps not. We really don't have any solid evidence

If the R1b/R1a split happening in a highly nomadic steppe population though, it may not be possible to pin it down geographically except in a very broad sense.

Question for anyone though - what do we know about mtDNA C1g found in our Karelian R1a friend?


@Onur -You're right. Comment withdrawn.

epoch2013 said...

One nice little detail that hasn't been noticed yet: Afontova Gora 3 has the same mtDNA as Hungarian Körös HG KO1: mtDNA R3 or R1b. Extremely rare nowadays.

FrankN said...

Ryan: On the Karelian C1g see here

http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0087612

Extinct, not documented in any other living or ancient population so far. Possibly related to extremely rare C1e that occasionally has been found among Icelanders. C1a is Siberian, C1b/c/d Amerindian. C1* has been documented in low frequencies across Europe (Iceland, Spain, Canaries, Germany, Norway) plus Lebanon and Bashkortistan.

I'd say, compared to mtDNA C1, figuring out the history of R1a/b is a piece of cake...

Grey said...

I think it's worth remembering that the mammoth steppe extended further than the current steppe.

The mammoth steppe extended in an arc all the way from Britain to Siberia so instead of western vs eastern HGs I think it's more likely to be southern vs northern and it wouldn't be surprising if people along the latitudinal border were mixed.

So maybe the model is mostly C WHG in the southern band, R1 in the northern mammoth steppe band and maybe IJ in the mountains in between.

The R1 people could have originated anywhere along the mammoth steppe eco niche and then spread to fill it.

Kristiina said...

Also the oldest mtDNA in Altai was R1b1* in this analysis from 2014
(http://dienekes.blogspot.lu/2014/06/ancient-dna-from-bronze-age-altai.html)

It is from the Bronze Age Altai (2742 BC) and yDNA is Q-M242.

It is said that he had probably brown eyes and he has black/dark brown hair.

Krefter said...

@Davidski,

Can you add a bunch of the new high coverage Paleo genomes to this spreadsheet.

https://drive.google.com/file/d/0B9o3EYTdM8lQSUlzbFpYZnlYaEU/view

epoch2013 said...

@Chad

"Loschbour may just be less Neandertal than Villabruna, causing it to be closer to ENA and ANE. David, can you check with that set you have with Neandertal?"

According to a 2012 John Hawks blogpost Ötzi has far more Neanderthal admixture than present day Europeans. While I can't seem to find good verification of this, it still might be interesting to see the result of this to see if we see similar things:

Mbuti Han Sardinian Iceman
Mbuti Han Anatolia_Neolithic Iceman

Mbuti AfontovaGora3 Anatolia_Neolithic Iceman

epoch2013 said...

@Kristiina

The paper on the Hungarian finds stated the only finds in present day persons were: one Finn, one Armenian (Family Tree database, http://www.familytreedna.com/) two Yakuts and one Bengali.

This too looks PIE or ANE related.

Davidski said...

epoch,

There might be minor effects here of attracion between ancient vs modern samples, due to deamination in the ancient samples.

Mbuti Han Sardinian Iceman_MN -0.0079 -2.991 959602
Mbuti AfontovaGora3 Sardinian Iceman_MN 0.0028 0.534 231532
Mbuti Han Anatolia_Neolithic Iceman_MN -0.0015 -0.591 951989
Mbuti AfontovaGora3 Anatolia_Neolithic Iceman_MN 0.0063 1.183 231491

Ryan,

You're not connecting the dots. Different subclades of both R1a and R1b have been found in prehistoric samples from very different sites in Eastern Europe, including those belonging to indigenous hunter-gatherers. So unless there were regular migrations from the real, as yet unidentified source of R1 differentiation into Eastern Europe, it looks like Eastern Europe was the main bifurcation region for R1, and the source of population movements carrying R1 subclaces to other parts of Eurasia. The latter view is backed up by genome-wide data.

ryukendo kendow said...

@ Davidski

How high quality are the Karasuk, Okunev and the Mezhovskaya samples in your dataset?

List coming later.

Davidski said...

A couple of the Karasuk are OK, but none of these samples are in my 1 million SNP dataset, so running them would drastically reduce the number of markers in each run.

Nirjhar007 said...

Things are actually quite boring , the new paper brought some surprises but I think its not of that scale that we wanted . Here we again see the same old boring and stalemate debates.

The other thing that I don't like is that, They Named the paper The genetic history of Ice Age Europe but didn't have any representative samples from Eastern Europe! , I think that is a huge hole and downer!.

So, I ask the leading authors that, what was the logic behind not having any samples from E Europe?.

Krefter said...

@Nirijhar007,

The famous Paleolithic remains are mostly from Western Europe. Paleolithic studies are more focused on Western Europe than anyother region. I have no idea why. It could be a domino affect. Someone found 30,000 year old cave paintings in Spain in 1865, then everyone wanted to study Paleolithic West Europe. We do have Kostinki12 and Kostinki14 and then we have the WHGs. So, we have the Paleolithic era and Mesolithic covered for most of Europe(Spain, France, Germany, Italy, Russia).

This is just my guess as to why ancient DNA studies have been focused on Europe and it may also be why Paleothic West Europe is focused on the most:

Archaeological studies are only a few hundred years old and done primary by Europeans. The foundation of modern pre-historic studies was set over 100 years ago and focused on Europe. When you hear phrases like Paleolithic humans, Neolithic spread of farming, Bronze age, etc., it's mostly only referring to Europe because that's where the archaeologist are from.

Even when studying of the ancient Egypt or Sumeria or whatever, there's focus on its influence on Europe. When they study history they mostly are focusing on their own history. I've noticed a consistent narrative to ancient human history which I think we get from European historians of the past. There's focus on The Origins of Civilization, and what they really mean is the origins of modern European civilizations.

For example; "cavemen" and "Cro Magnon" are just early humans from Europe. Yet, our culture views them as THE early humans and not even human but human-like apes. They're viewed as THE early humans and the ancestors of all of humanity, because European pre-historians are mostly concerned about European pre-history. Cro Magnon are the earliest Europeans but not earliest humans.

Genetics like David Reich have to look back at archaeology to chose what ancient DNA samples they get. For the reasons listed above, they like archaeologist are going to be mostly focused on European pre-history(which is viewed as generic human pre-history).

FrankN said...

Here is some background on Afontonova Gora, for anybody interested:

http://www.donsmaps.com/afontovagora.html

Riverside location (upper Yenisei). Spear straighteners made from deear antlers. Carefully rounded stones that may have served as bolas, or net sinkers.
Let me put it like this: I wouldn't focus too much on mammooths, the subsistence base appears to rather have been ungulates (deer) and fish. Rather than "the Steppe", watercourses seem to have been the main communication channels. Several other UP East European sites, e.g. Kostenki on the Don, have a similar location as AG, namely uphill overseeing a major river.

The Yenisei location is intriguing. The mammooth maps linked by epoch further above show evidence of pines on the lower Yenisei and along the Siberian Polar Sea coast 15-20 kya. The Polar Sea seems to have been relatively ice free during interstadials, and might have served as communication channel into Karelia, as well as North East America, where R1b(1) is the predominant Amerindian yDNA....

Davidski said...

There's no native R1b in the Americas. It's all from the colonial period.

ryukendo kendow said...

@ Davidski

David, we could check them against Siberians, Amerinds and East Asians. They allow us to 'plumb the depths', so to speak, of the multiple layers of Amerind-like, Chuckhi-Koryak-like and East Asian-like ancestry in Siberia, which is a rich seam we haven't mined yet. How's the overlap?

Nick has replied to me on Razib's place as well, apparently ADMIXTUREGRAPH is finally coming out. Hope you could use it, nothing better to separate the rivulets in the braided streams we are often dealing with.

Davidski said...

Okunevo would be interesting, but the overlap is poor and deamination effects high, so there's not much point.

Karasuk and Mezhovskaya are fairly recent samples with complex ancestry, and can probably be easily replaced by modern samples like Altaians, Chuvashs and Erzya.

Davidski said...

Ah, maybe in that case just wait for qpGraph. Treemix is pissing me off a bit.

Hopefully when that new version comes out it's easy enough to install.

ryukendo kendow said...

@ David
David, I mean could we have stats of the form
IE/LNBA Okunevo/Karasuk/Mezhovskaya ENA1 ENA2
Not that the Eastern-admixed IE groups are of any interest in and of themselves in Treemix or so on.
For this even poor marker overlap is still OK. I just wanted to know how good it would be.

Davidski said...

Yeah, I can run some D-stats with them.

Karl_K said...

@epoch2013

Further analysis shows that the Iceman genome contains levels of archaic ancestry that are comparable to that of other non-African populations.

ryukendo kendow said...

Davidski, please provide me with the labels in the dataset with many more asian groups, the one with 500k markers. I'll give you back a .txt file.

epoch2013 said...

@Karl_K

I guess that simply had to be. How could such an old admixture have a far higher level in a population that clearly is related to him but hasn't got that higher level? It would have puzzled me.

You got a link to a page or article on that further analysis? Or do I simply need to search John Hawks blog?

Gioiello said...

@ FtrankN
"The Polar Sea seems to have been relatively ice free during interstadials, and might have served as communication channel into Karelia, as well as North East America, where R1b(1) is the predominant Amerindian yDNA".

Interesting to know that there is an R1b Native American. There are some new data that demonstrate that?
When I checked this question many years ago I din't find in Native Americans any old R1b subclade except an R1b1 of the type from the Isles I thought introgressed in native Americam from the European colonists...

I am seeing that Davidski doesn't think as you:
"There's no native R1b in the Americas. It's all from the colonial period".



Davidski said...

rk,

https://drive.google.com/file/d/0B9o3EYTdM8lQMnFaTElMWDBaVFU/view?usp=sharing

I also have Paniyas and Kets with over 500K SNPs in my European dataset, but that one is missing the Human Origins samples, Okunevo, Karasuk, and a few other Allentoft et al. samples.

Karl_K said...

@epoch2013

http://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1004353

Nirjhar007 said...

Krefter,

Please type my name correctly. My question was not of what you answered, my question was, how they can give the paper such name without having samples from areas of North of Caspian,samara, ,Volga-Don etc .

And to really find the truth about Davids R1 bifurcation suggestion in E Europe , one must test those areas samples dating from 7th millennium BC at least.

According to my theory there shouldn't be much R1a/R1b in E Europe before 6000 BC . But we have Karelia , but I am suspicious that it wasn't a widespread group by its time.

The Hotu and Belt Cave Y-dna of Iran will be a good candidate for the source population of R1a in Europe I think .

I am also getting suspicious of how R1b-M269+ can be considered original PIE or not , I am beginning to think that It was R1a-M417 that fits better with some specific groups of J2 as well.

Karl_K said...

@Gioiello FrankN

"Interesting to know that there is an R1b Native American. There are some new data that demonstrate that?"

If there were truly native Native American R1b haplogroups, it would be a huge story.

But all of the papers looking at Native American haplogroups set all of the R1b samples aside because they are extremely very obviously from western Europeans.

The same is true in South America and North America. Most of them in South America cluster with the Spanish.

It would be very interesting to see where the North American ones cluster, but genetic testing is only now becoming less taboo for native North Americans. Maybe the Kennewick Man results will spur more of them to get genetic testing.



Jaydeepsinh Rathod said...

Epoch & Chad,

Thank you for your inputs. So SC Asia can be a possible place of origin of R1b . I looked into the Iranian y-dna paper. It is amazing to see how much of the y-dna profile of the Iranians is similar to the Indians. The Iranians share with the Indians the following Ydna HGs :- R1a, R1b, R2a, J1, J2, L, H, Q, G & T. It appears that compared to other populations, the Iranians & SC Asians should form a close group.

David,

Thank you for detailed response.

"So if you posit that one or a couple of them originated in South Central Asia, then you're basically implying that they all did."

The early Europeans might share drift with South Asians in comparison with East Asians because of their shared origin. However, the later Europeans like Villabruna, if there was no migration between them & SC Asians, should grow more distant to SC Asians as compared to populations like Kostenki who date close to the time of West Eurasian split. If on the other hand Villabruna shares more drift with SC Asians in comparison to Vestonice, ElMiron & Kostenki, then it should mean some sort of admixture - either from SC Asian to Villabruna or vice-versa or into both of them from a 3rd source. Atleast thats how I understand it.

"I suppose you could argue that South Asia was like a reservoir of this West Eurasian super clade, and its different offshoots moved north into Europe, the Near East and Siberia every few thousand years. But as we get more samples from Europe and northern Asia, the gaps fill, and it's already pretty clear that there's no need to look to South Asia to find any relatively recent ancestral groups, because they're all there in Europe and nearby going back to the LGM at least."

I think the need to look into South Asia arises simply to get a better picture of the whole scenario. Just looking at evidence from Europe & Northern Asia cannot give us a complete picture of the origins & migrations. We need to look at evidence from further south.



Jaydeepsinh Rathod said...

Gioiello,

Thnaks for all the info. The younger R1b clades in South Asia probably have a relationship with the Yamnaya people because I believe we have the Yamnaya Z2013 in South Asia. Correct me if I am wrong. As for the presence of older subclades of R1b, they probably have been present in SC Asia since the pre-Neolithic.

Sein,

" I think the reason South Central Asians have the highest amounts of ANE admixture out of all predominantly West Eurasian populations, is primarily due to them deriving their genetic heritage from Andronovo-like steppe populations (which had a lot of ANE, more than even the most ANE-admixed northern/eastern Europeans), CHG-like West Asian agriculturalists (who seem to have been predominately ANE, with Basal Eurasian admixture, and probably some Villabruna-related admixture), and South Asian hunter-gatherers (who seem to have had, based on evidence from contemporary populations, some kind of component related to ANE, perhaps a very basal branch in the North Eurasian group). So in theory, all three "ancestral populations" (to use cookie cutter, overly sloppy/simplistic language, for the sake of brevity) for South Central Asians had very substantial amounts of ANE ancestry."

You are positing two migrations in South Asia when we do not have evidence of even one of these migrations! We should remember that at the end of the day, all of these are hypothetical and may never have happened. You do admit that atleast some of ANE in South Asia is indigenous. The CHG too should have had its presence in SC Asia since the pre-Neolithic, because, as I have said, the presence of J2 in SC Asia appears to be older and cannot be related to the Neolithic expansion from the Near East.

An expansion of CHG from Caucasus into SC Asia should bring the basal Eurasian ancestry. Do we have basal Eurasian in SC Asia ? I believe there is negligible presence of ENF component in SC Asia and that is related to historical Islamic expansions. On the other hand, a steppe migration into SC Asia should show evidence of EHG component in SC Asia. Do we have that ? I think the situation is complex and cannot simply be explained by conventional theories.

Nirjhar007 said...

Jaydeep, the R1b in south Asia is nothing of any structure. Quite like the case of some occasional Z-93's in Europe Today .

Karl_K said...

@Nirjhar007

"R1b in south Asia is nothing of any structure. Quite like the case of some occasional Z-93's in Europe Today"

A bunch of migrants from India!

Onur said...

The R-Z93 subclades of Europe/West Asia and those of South Asia are different. They parted ways already during the early to middle Bronze Age. But both seem to be ultimately rooted in the Corded Ware culture of Eastern and Central Europe.

epoch2013 said...

Theory: If the idea that the increasing affinity to Han, ME and American is due to the decreasing amount of Neanderthal then maybe it would somewhat like this: A number of genes for which all unadmixted humans are ancestral are derived in Neanderthals. So humans with Neanderthal admixture have derived genes for which Mbuti has ancestral genes. If the amount of Neanderthal genes diminish over time until we get to WHG, then WHG will share more ancestral genes with Mbuti than with earlier UP Europeans and Mbuti would show as much of a signal as Han, American and Middle-Eastern.

There is this in the paper:

“In both models of Figure S6.4, Villabruna, too, is inferred to be admixed, with 1-2% of its ancestry deriving from a deep Eurasian branch that split before the separation of UstIshim from all other Eura
sians (drift distance ~0.09). This lineage is inferred to derive from earlier than the founder of all non-Africans, and is inferred to be more drifted than the “Basal Eurasian” lineage.”


Would this show some results?

Chimp Mbuti Hungary_HG Muierii2
Chimp Mbuti Hungary_HG Paglicci133
Chimp Mbuti Hungary_HG Pavlov1
Chimp Mbuti Hungary_HG Vestonice16

FrankN said...

Re: Amerindian R1b

I am not aware of any recent studies on Amerindian R1b. However, in the following study of pacific NW Amerindians, two inland populations (Splatsin and Stswecem'c, both part of Salishan-speaking Shuswap people) cluster close to, or among Europeans. This is interpreted as reflecting recent admixture. However, ADMIXTURE runs place them in a separate, orange-coloured cluster at K=10, while in other Amerindian populations, including MXL, the European signals (light and full blue) are maintained and not replaced by that orange component (Fig. 8).
http://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1004530#s5

Another recent study on the coastal Tsimshian finds no evidence of European admix in Tsimshian aDNA, but, however, an respective signal in Anzick-1 (Fig. 1a). Current European admix in Tsimshian is estimated at some 33%. "The best-fitting model suggests that a bottleneck occurred approximately 175 years BP (bootstrap 95% confidence interval: 125-225, Table 1) in the ancestors of the modern Tsimshian with an accompanying reduction in effective population size
of 57%. The timing of the bottleneck coincides with the documented smallpox epidemics of the 19th Century and historical reports of large scale population declines. A majority of the European admixture in the population likely occurred after the epidemics."

http://biorxiv.org/content/biorxiv/early/2016/04/29/051078.full.pdf

This gives us some baseline on the European admix that may be expected in smaller, isolated communities with relatively late European contact. Another baseline, for a large population with early European contact, is provided by the Maya, with some 10% European admixture. In relation to both, R1(b) percentages above 50%, as reported for Chippewa (Ojibwe), Seminole, Sioux and Cherokee look extraordinarily high. Note here also that IIRC both the Raghavan and Skoglund/Reich 2015 papers on the Peopling of the Americas failed to identify Athabascans that weren't signficantly "European admixed".

There are two scenarios that might explain such high R1(b) scenarios from post-columbian contact:

1. Early admix (fur trade), with founder effects that were enhanced by better resistance to European diseases (smallpox etc.): This scenario has been refuted for the Tsimshian (see above). It also doesn't apply to the Ojibwe, where Bolnick e.a. failed to identify any star-like structure of R1(b) lineages, but instead more than 60 different R1 haplotypes among 80 R1 individuals:
http://mbe.oxfordjournals.org/content/23/11/2161.full.pdf

2. Late (19-20th century) admix from various sources: In that case we would expect:

(i) A distribution of "European" yDNA among Amerindians that more or less mirrors the genetic profile of European settlers, i.e. includes substantial I, R1a, etc. However, Hammer e.a. (Fig. 1) provide the following rations for the USA (Native American share, European Ancestry share, quotient), which disprove that expectation (and also speak against Vikings/Icelanders as a putative source of pre-Columbian R1b ingression):

R1b1a2: 21.9%/58.3% = 0.38

R1a1a: 1.5%/7.2% = 0.21
I1(P30): 2.8%/11.7% = 0.24
I2a1 (P37-2): 0.8%/2.7% = 0.30
I-P19* (I2a2?): 1.0%/4.9% = 0.20
G2a (P15): 0.3%/3.6% = 0.08
Subtotal: 6.4%/30.1% = 0.21

https://www.familytreedna.com/pdf/HammerFSIinpress.pdf

(ii) Amerindian R1(b) haplotypes that are identical to European ones. However, Bolnick e.a. (Fig. 6, see above) reported 20 out of ~60 Amerindian haplotypes not shared with Europeans, five of which were 4-8 mustations removed from European ones. [Goiello: You said you checked Amerindian subclades some years ago - did that include the ones from Bolnick e.a.?]

In short: I have little doubt that a good part, maybe 2/3, of Amerindian R1b relates to recent European admixture. But there is quite some indication of an older admixture layer that appears to have received little attention in research so far.

Gioiello said...


@Jaydeepsinh Rathod

"Gioiello,
Thnaks for all the info. The younger R1b clades in South Asia probably have a relationship with the Yamnaya people because I believe we have the Yamnaya Z2013 in South Asia. Correct me if I am wrong. As for the presence of older subclades of R1b, they probably have been present in SC Asia since the pre-Neolithic".

Of course my theory, based only upon the study of the Y and the mt, wanted to demonstrate that the subclades from R-L389+ came from Western Europe and not from Middle East or anywhere. Of course I think that the R-L278+/L389- haplotypes present in India and Asia are very old there, because they separated as to Yfull 17200 years ago and after they have two different descents. The Indian haplotypes seem having come from Central Asia/Caucasus, thus we may think that the common ancestor were in Southern Siberia. My theory of the Italian Refugium didn't say anymore that hg. R was born in Western Europe, but I said that it was in the Italian Refugium at least from the Younger Dryas. Villabruna sample of 14000 years ago seems to demonstrate that it was here also before. We don't know so far since.
But I agree with you all that only many other aDNA from everywhere could give us a more reliable picture. But from Villabruna to R-M269 and also R-M73 (amongst my theories also this time against all there was also the origin of M73 in Western Europe and not in Asia even though it is more diffused there now) to R-L23 to R-L51 etc, I expect that aDNA demonstrates all that.


Karl_K said...

@FrankN

"In short: I have little doubt that a good part, maybe 2/3, of Amerindian R1b relates to recent European admixture. But there is quite some indication of an older admixture layer that appears to have received little attention in research so far."

As I said before, I think this would be a very interesting area of investigation. However, it is almost certain that 100% of these R1b haplogroups areived after the year 1492.

I know there is a paper about Y haplogroups in Greenland, and they also found lots of typical European R1b among the natives Greenlandians.

Rob said...

Onur

"The R-Z93 subclades of Europe/West Asia and those of South Asia are different. They parted ways already during the early to middle Bronze Age. But both seem to be ultimately rooted in the Corded Ware culture of Eastern and Central Europe"

Hold on, Z93 in Europe today is overwhelmingly found in Ashkenazi & Roma men. So they can't have been in Europe since the Bronze Age

Krefter said...

@Davidski,

Can you add a bunch of the new high coverage Paleo genomes to this spreadsheet.

https://drive.google.com/file/d/0B9o3EYTdM8lQSUlzbFpYZnlYaEU/view

Davidski said...

@Rob

Hold on, Z93 in Europe today is overwhelmingly found in Ashkenazi & Roma men. So they can't have been in Europe since the Bronze Age.

Overwhelmingly, but not entirely, because there are basal lineages of Z93 in Poland and Russia that are not recently related to Asian Z93.

Quite apart from that, I'm pretty sure Poltavka and Srubnaya people weren't Jews or Roma.

@Krefter

I'll look into it. I might have to do a new sheet.

Davidski said...

Jaydeepsinh Rathod,

Formal stats support the idea that West Eurasian-specific drift is intrusive in South Asia, and probably arrived there in multiple waves, including relatively recently via Central Asia.

Even within the same regions, like Gujarat, there are still clear differences in this context among different communities.

Mbuti Tajik_Shugnan Ust_Ishim Kostenki14 0.0422 8.6
Mbuti Kalash Ust_Ishim Kostenki14 0.0348 8.055
Mbuti Pathan Ust_Ishim Kostenki14 0.0315 7.528
Mbuti GujaratiA Ust_Ishim Kostenki14 0.0312 6.923
Mbuti GujaratiB Ust_Ishim Kostenki14 0.0224 4.931
Mbuti GujaratiC Ust_Ishim Kostenki14 0.0212 4.627
Mbuti GujaratiD Ust_Ishim Kostenki14 0.0199 4.302

Mbuti Tajik_Shugnan Kostenki14 Villabruna 0.0391 8.135
Mbuti Kalash Kostenki14 Villabruna 0.0316 7.607
Mbuti Pathan Kostenki14 Villabruna 0.0308 7.725
Mbuti GujaratiA Kostenki14 Villabruna 0.028 6.424
Mbuti GujaratiB Kostenki14 Villabruna 0.0229 5.147
Mbuti GujaratiC Kostenki14 Villabruna 0.0239 5.439
Mbuti GujaratiD Kostenki14 Villabruna 0.0193 4.35

epoch,

Chimp Mbuti Hungary_HG Muierii2 -0.0272 -2.402
Chimp Mbuti Hungary_HG Paglicci133 -0.0273 -2.329
Chimp Mbuti Hungary_HG Pavlov1 -0.0097 -0.695
Chimp Mbuti Hungary_HG Vestonice16 -0.017 -3.736

Seinundzeit said...

David,

If you do create a new sheet, could these be the outgroups (based on the best sheet you did, with some new samples)?

Ami
Afontova Gora2
Anatolia_Neolithic2
Australian
BedouinB2
Biaka
Caucasus_HG2
Dai2
Eskimo_Naukan
Han
Iberia_Chalcolithic
Iberia_Mesolithic
Karitiana
Kharia2
Kostenki12 (or Kostenki14)
MA1
Mansi2
Motala_HG
Ostuni1
Oase1
Onge2
Papuan2
Samara_HG
Selkup
Yoruba

And the new samples to be included as test populations for modelling/fits would be:

Afontova Gora3

GoyetQ116-1

Vestonice16

El Miron

Villabruna

Kostenki14 (or Kostenki12)

And split HGDP00213, HGDP00218, and HGDP00224 as "PathanB", and the rest of the Pashtun samples as "Pathan".

This would be pretty amazing, some great stuff could be done with this sheet. Of course, only when you get the time, and inclination, to do this.

Also, could you eventually try these stats:

Mbuti Tajik_Shugnan Ust_Ishim AfontovaGora3
Mbuti Kalash Ust_Ishim AfontovaGora3
Mbuti Pathan Ust_Ishim AfontovaGora3
Mbuti GujaratiA Ust_Ishim AfontovaGora3
Mbuti GujaratiB Ust_Ishim AfontovaGora3
Mbuti GujaratiC Ust_Ishim AfontovaGora3
Mbuti GujaratiD Ust_Ishim AfontovaGora3

Mbuti Tajik_Shugnan Ust_Ishim MA1
Mbuti Kalash Ust_Ishim MA1
Mbuti Pathan Ust_Ishim MA1
Mbuti GujaratiA Ust_Ishim MA1
Mbuti GujaratiB Ust_Ishim MA1
Mbuti GujaratiC Ust_Ishim MA1
Mbuti GujaratiD Ust_Ishim MA1

Mbuti Tajik_Shugnan Kostenki14 AfontovaGora3
Mbuti Kalash Kostenki14 AfontovaGora3
Mbuti Pathan Kostenki14 AfontovaGora3
Mbuti GujaratiA Kostenki14 AfontovaGora3
Mbuti GujaratiB Kostenki14 AfontovaGora3
Mbuti GujaratiC Kostenki14 AfontovaGora3
Mbuti GujaratiD Kostenki14 AfontovaGora3

Mbuti Tajik_Shugnan Villabruna AfontovaGora3
Mbuti Kalash Villabruna AfontovaGora3
Mbuti Pathan Villabruna AfontovaGora3
Mbuti GujaratiA Villabruna AfontovaGora3
Mbuti GujaratiB Villabruna AfontovaGora3
Mbuti GujaratiC Villabruna AfontovaGora3
Mbuti GujaratiD Villabruna AfontovaGora3

This would be greatly appreciated. Thanks in advance.

Davidski said...

Sein,

Some of those outgroups have less than 100K SNPs, so they probably wouldn't produce stable results. Anywho...

Mbuti Tajik_Shugnan Ust_Ishim AfontovaGora3 0.0839 14.905
Mbuti Kalash Ust_Ishim AfontovaGora3 0.0777 15.442
Mbuti Pathan Ust_Ishim AfontovaGora3 0.0715 15.162
Mbuti GujaratiA Ust_Ishim AfontovaGora3 0.0682 13.07
Mbuti GujaratiB Ust_Ishim AfontovaGora3 0.0612 11.771
Mbuti GujaratiC Ust_Ishim AfontovaGora3 0.0582 11.455
Mbuti GujaratiD Ust_Ishim AfontovaGora3 0.0535 10.133

Mbuti Tajik_Shugnan Ust_Ishim MA1 0.0813 16.277
Mbuti Kalash Ust_Ishim MA1 0.0759 16.662
Mbuti Pathan Ust_Ishim MA1 0.0697 16.434
Mbuti GujaratiA Ust_Ishim MA1 0.0675 14.71
Mbuti GujaratiB Ust_Ishim MA1 0.0581 12.355
Mbuti GujaratiC Ust_Ishim MA1 0.0588 13.014
Mbuti GujaratiD Ust_Ishim MA1 0.0531 11.191

Mbuti Tajik_Shugnan Kostenki14 AfontovaGora3 0.0445 7.299
Mbuti Kalash Kostenki14 AfontovaGora3 0.0485 8.849
Mbuti Pathan Kostenki14 AfontovaGora3 0.0448 8.658
Mbuti GujaratiA Kostenki14 AfontovaGora3 0.0421 7.748
Mbuti GujaratiB Kostenki14 AfontovaGora3 0.0432 7.667
Mbuti GujaratiC Kostenki14 AfontovaGora3 0.0441 7.994
Mbuti GujaratiD Kostenki14 AfontovaGora3 0.0374 6.547

Mbuti Tajik_Shugnan Villabruna AfontovaGora3 0.0049 0.849
Mbuti Kalash Villabruna AfontovaGora3 0.0145 2.722
Mbuti Pathan Villabruna AfontovaGora3 0.0118 2.467
Mbuti GujaratiA Villabruna AfontovaGora3 0.0123 2.296
Mbuti GujaratiB Villabruna AfontovaGora3 0.0167 3.189
Mbuti GujaratiC Villabruna AfontovaGora3 0.0179 3.417
Mbuti GujaratiD Villabruna AfontovaGora3 0.0166 3.077

Chad Rohlfsen said...

A little off-topic, but I wanted to pass along models without Basal Eurasian work on qpAdm, and the same thing shows up on TreeMix.

Basically, first there is an Ust-Ishim people stretching from Europe through Siberia, as Oase1 is only Ust_Ishim with more Neandertal. Next, we have Aurignacians and Gravettians that may ultimately stem from West Asia. Using Iberia EN as this West Eurasian pop gives decent fits. Just in case people forgot, Iberia EN fails as a mix of Anatolian and WHG and also forms the extreme pole of West Eurasians in a global PCA. I can share some of these fits to show that Basal Eurasian is not needed to explain the phenomenon. Crazier yet is that CHG is a good fit as BedouinB plus WHG, which lines up perfectly on the global PCA. It's all pretty wild, but interesting none-the-less.

ryukendo kendow said...

@ Jaydeep

Jaydeep, people here have already done many, many analyses which show unambiguously that all Indians, incl South Indians, have large proportions of Kotias ancestry, and all carry levels of Basal Eurasian ancestry. Furthermore, High-caste and Low-caste populations from the same region are strongly differentiated by their separate levels of similarity to Northeast Europeans and Indo European aDNA from the Steppe on one hand, vs Middle Easterners on the other; for example, GujaratiD is closer to Middle Easterners, while GujaratiA is closer to European populations and LNBA genomes from the Steppe; likewise Lodhi is closer to Middle Easterners/CHG, and Tiwari to Europeans.

There is some possibility that J2 or CHG may have been present in India even before the Neolithic; the CHG genomes were found in a Hunter-gatherer context after all. But what is undeniable is that a cline of CHG/Middle-Eastern-like ancestry, plus a cline of NE-European-like ancestry, exists in S Asia, the former more associated with geography than caste, while the latter is more strongly associated with caste.

Whatever the origins of R as a whole, a movement of European-like autosomes into India, carrying R1a, that affected the ancestors of the various castes differently, is probably undeniable at this point.

@ David
David, the post containing the plots with GujaratiA and GujaratiD int the comment sections in the convo w Alberto seems to be deleted, or has it? @Anyone, has a copy of it? @Alberto, do you?

Chad Rohlfsen said...

First, I had to make Oase1 as a mix of Ust_Ishim and Chimp as this set doesn't have Neandertal, but David could try it with his Human Origins set.

Oase1 comes out 93.4% Ust_Ishim and 6.6% Chimp
std errors are 0.013 for both and the chi-square is 1.296 with a tail of 0.971862

ryukendo kendow said...

@ Chad

Chad, for those qpAdm fits, did you include Ust Ishim in the ougroups? If you did not, could you try them with U-I there?

I have a suspicion about whats going on.

ryukendo kendow said...

I mean the fits with Paleolithic europeans fitted as Iberia EN.

Davidski said...

rk,

Which plots with GujaratiA and GujaratiD? PCA based on D-stats or genome-wide data?

Chad Rohlfsen said...

No, they have Ust Ishim or Oase1 as a component. It just depends on the Neandertal level of the sample.

Gioiello said...

@ FrankN

(ii) Amerindian R1(b) haplotypes that are identical to European ones. However, Bolnick e.a. (Fig. 6, see above) reported 20 out of ~60 Amerindian haplotypes not shared with Europeans, five of which were 4-8 mutations removed from European ones. [Gioiello: You said you checked Amerindian subclades some years ago - did that include the ones from Bolnick e.a.?]


You may look at this interesting discussion on www.eng.molgen.org,
R1b Found in Native American Populations, but those Native American samples I examined very likely on www.worldfamilies.net.

Seinundzeit said...

David,

Very fascinating, thanks for running these!

South Central Asians/South Asians seem significantly more related to Afontova Gora3 than they are to Kostenki14, and this still holds to a lesser extent in the Villabruna vs Afontova Gora3 comparison (which is expected, assuming Villabruna has some ANE admixture).

Chad Rohlfsen said...

Kostenki14
42.6% Ust_Ishim
15.1% AG3
42.4% Iberia_EN

std errors .137 .043 .163

chi-square 3.105 tail 0.683832

Chad Rohlfsen said...

Villabruna

16.6% Oase1
19.2% AG3
57% Iberia_EN
7.1% Han

std errors .342 .110 .448 .026

chi-square 5.163 tail 0.271013


11.6% Ust-Ishim
13.4% AG3
70.2% Iberia_EN
4.9% Han

std errors .126 .041 .143 .019

chi-square 2.520 tail 0.641052

Chad Rohlfsen said...

OOOOhhhhh!

Villabruna

16.4% Ust_Ishim
19.6% MA1
62.8% Iberia_EN
1.2% Han

std errors .102 .041 .115 .017

chi-square 1.090 tail 0.895914

Davidski said...

So what's Iberia EN in your opinion? Descendants of southern Balkan hunter-gatherers, or migrants from the Levant or nearby?

Chad Rohlfsen said...

Well, here is another model that requires no ENA into WHG to explain the stats.

19.6% Ust_Ishim
20.5% MA1
59.8% Iberia_EN

std error .093 .039 .109

chi-square 1.935 tail 0.925609

Mbuti Han Kostenki14 Ust_Ishim 0.0078 1.616 52273 51463 906084
Mbuti Han Iberia_EN Ust_Ishim 0.0114 2.893 51137 49984 873267
Mbuti Han Iberia_EN Kostenki14 0.0037 1.008 46717 46375 839112
Mbuti Han Ust_Ishim MA1 0.0183 3.828 40059 38616 683873

There is interesting stuff here that kind of agrees with the models. Kostenki looks between Iberia EN and Ust Ishim with some MA1. Also, this shows that since WHG is very significantly different in Ust_Ishim vs MA1, that is enough to give them an ENA shift. If Basal Eurasian is real, why is Iberia EN the same distance from East Asians as Kostenki? Loschbour can be closer to ENA than Iberia HG due to being 40% MA1 and Ust_Ishim. Iberia EN as a proxy as pure West Eurasian is further from ENA than everyone, and also closer to Africans than all without definite SSA. I also wonder if this means West Eurasian broke off first, then ENA after an intermediary step in Ust_Ishim, which may be also basal to ENA, which is why it is closer to East Asians. K2 ydna... You never know. It is simplistic, but makes sense. Global PCAs just don't reflect anything like Basal Eurasian being needed and doesn't require so many extinctions, but just everyone being a mix, of a mix, of a mix. It could be possible that BE is real, but it could also be that Iberia_EN is the closest we have to another hunter that is the significant part of the ancestry of WHG, minor in UP, and heavy in the Near East, so the model works without being totally true, due to lack of a real reference pop. What do you think, David? It's funny how this kind of mimics TreeMix, with WHG off the Anatolian CHG group and Ust_Ishim / Oase1 into UP. There is also that 6% African into the root of that group. I can try that fit.

Chad Rohlfsen said...

I meant Ust_Ishim and K2 basal to a good chunk of ENA and ANE, which may explain why they are significantly closer to each other than Iberia EN is to E Asians. If they branch off the same tree but split super early, that might explain it. Oh, also Iberia HG is about the most significantly shifted away from Ust_Ishim of any non SSA mixed pop. Which is also explained by the global PCA, without a BE model. I'm trying to figure out how ANE can be modeled. It's a pain in the ass.

Chad Rohlfsen said...

Iberia_EN not HG. ^^

ryukendo kendow said...

@ Chad
Chad, could you please repeat the Oase runs with Ust Ishim as an outgroup, and Ust Ishim runs with Oase as an outgroup? Because the modern outgroups will be simply unable to differentiate between old West Eurasian HG variation of the Kostenki Type, pulling away from Han, and Basal Eurasian variation of the Kotias-EEF type, which also does that. We know for a fact that Basal Eurasian in kotias and Satsurblia simply cannot explain the 'Basal' in UP Europeans, because Kotias and Satsurblia favour each other strongly over Kostenki despite all three being close to equidistant from Han; so I find it quite likely that UP Europeans being fit as EEF+ANE is due to a perspective/dimensionality error due to not having the right outgroups.

We've seen this before. Papuan as ~90% Ami + ~10% Denisovan works very well, until you include a west Eurasian outgroup, which causes the whole thing to fall through.

Chad Rohlfsen said...

I don't know, but saying that everyone that lived over 17k years ago just died off doesn't make sense when tool industries continue. This would also connect Gravettian and Mal'ta with shared ANE, or whatever kind of mix that is, or is shared between them. Whether it's Ust Ishim, which we see is basically identical to Oase1. That would sure connect Siberia and Europe.

Chad Rohlfsen said...

Another funny thing is that Villabruna/WHG and Satsurblia basically form a clade compared to the UP samples on the global PCA. It's funny because both are epiGravettian. IJ brothers with an occasional R1 spatter from ANE.

Chad Rohlfsen said...

Ryu,

I don't think you could have Ust_Ishim and Oase1 in left and right. If they're identical it wont work. It's like modeling someone with Loschbour and LaBrana as an outgroup. I'll give it a shot though.

ryukendo kendow said...

@ Chad

Maybe you are correct, but we have no way of knowing until we see how the fits look like when either Oase or Ust-Ishim is translocated to the other side to the outgroups. Could you try these fits? For the Villabruna fits, you could translocate Kotias to the other side to see as well, since Kotias and Iberia_EN would be separated from >25KYA. If the fits do not decline, who knows, maybe you're right!

However, I'm quite sure that at least one of the fits above is caused by dimensionality error.
http://www.nature.com/nature/journal/v524/n7564/images/nature14558-st1.jpg
We see that Outgroup Oase Eurasian1 Eurasian2 is always insignificant, even when Ust Ishim is one of Eurasian1 or Eurasian2. This shows very clearly that Oase does not demonstrate affinities to any later Eurasian, not even Ust-Ishim. Neanderthal in Oase should not affect this stat as Oase is beside the outgroup, so all the excess neanderthal is on one side without confounds.

So this:
"Oase1 comes out 93.4% Ust_Ishim and 6.6% Chimp
std errors are 0.013 for both and the chi-square is 1.296 with a tail of 0.971862"

Is almost certainly due to outgroups not capturing dimensionality where Oase and Ust-Ishim are highly different, which we can't really do anything about at the moment cause we have no population that is differentially related to Oase and Ust-Ishim not due to Neanderthal admix to plonk in the outgroups.


@ David
David, I did a plot of GujaratiA and GujaratiD with stats you provided, during a convo w Alberto, I was hoping someone had saved it.

Seinundzeit said...

I used to think that "Basal Eurasian" as described in Lazaridis et al. was an unnecessary construct, in light of K14's relationship to ENA. But now, I'm pretty sure it's real.

If we assume that Villabruna/Loschbour/La Brana and Neolithic Anatolians/EEF are closely related (which is a pretty reasonable assumption), but differentiated by some sort of admixture not shared by both, then we must assume the existence of a cluster of "Basal Eurasian" populations contributing ancestry to Neolithic Anatolians/EEF, but not to Villabruna/Loschbour/La Brana. It's absolutely necessary.

The only other option would be having huge amounts of ENA and MA1/Afontova Gora-related admixture in Villabruna/Loschbour/La Brana, to the exclusion of Neolithic Anatolians/EEF. But that still wouldn't add up. Many stats could never be explained in such a model.

Chad Rohlfsen said...

They're too identical. The chi and the tail aren't great. Let me try my leftpop that separated the two, Chimp. I'll see if that cleans it up.

ryukendo kendow said...

@ Chad

Chad, when all the members of the outgroup are close to being equally related to Ust Ishim and Oase and Chimp, but are all a step lower in relations to Oase than Ust-Ishim, and a huge cliff lower in relation to Chimp, then it makes perfect sense to model Oase as Chimp+Ust Ishim, which may even work with great accuracy. The algorithm wouldn't know any better, I would think. To disprove this we would need to 'inpute' in information into the program that Oase is also equally related to Ust-Ishim and all the outgroups; however the program requires we place Oase in the outgroups for that piece of information to register, and we can't have him in both the ingroups and outgroups.

As it stands, I think the figs in the Nature img give us an extra datapoint which immediately changes our interpretation of that fit, since we know for a fact that the info in the program is incomplete no matter how we arrange Ust-Ishim and Oase.

Chad Rohlfsen said...

What I'm getting here is MA1 into WHG to the exclusion of all EN pops. ENA isn't necessary if you look at the MA1 stats. Having Ust-Ishim and especially MA1 would pull someone closer to ENA than Anatolians and other farmers.

Davidski said...

rk,

Your GujaratiA vs GujaratiD plots are here.

http://i.imgur.com/5stDC2E.png

http://i.imgur.com/5stDC2E.png

(European blue, Middle Eastern red, ENA-affected green)

ryukendo kendow said...


Wow thanks! Where did you find them?
@ Jaydeep

Look at these that David posted, they tell us a great deal.

Chad Rohlfsen said...

Mbuti Oase1 Ust_Ishim Han 0.0077 1.170 9217 9075 170489
Mbuti Oase1 Ust_Ishim Yoruba -0.2532 -43.285 7752 13008 170489
Mbuti Oase1 Ust_Ishim Karitiana 0.0092 1.312 9248 9079 170489
Mbuti Oase1 Ust_Ishim Ju_hoan_North -0.3070 -51.096 7375 13909 170487
Mbuti Oase1 Ust_Ishim Papuan 0.0023 0.330 9118 9076 170488
Mbuti Oase1 Ust_Ishim Japanese 0.0084 1.272 9230 9076 170368
Mbuti Oase1 Ust_Ishim Biaka -0.3039 -52.493 7380 13823 170489
Mbuti Oase1 Ust_Ishim Dai 0.0085 1.297 9209 9054 170489

Chad Rohlfsen said...

With Kotias added to the right pop

Villabruna

7.7% Ust_Ishim
17.9% MA1
72.9% Iberia_HG
1.5% Han

std errors .051 039 .057 .017

chi-square 1.696 0.791378

Chad Rohlfsen said...

Iberia_EN, sorry ^^^^

ryukendo kendow said...

Did this have Oase in the right as well? We most definitely can include it, as the previous set of stats show that Oase is at least as far away from all other pops as Ust-Ishim is.

«Oldest ‹Older   1 – 200 of 227   Newer› Newest»