search this blog

Thursday, November 13, 2014

TreeMix graphs with Kostenki14 and Ust'-Ishim


First of all, here's a map with some basic info about these two Upper Paleolithic North Eurasian genomes. They're separated by less than 10,000 years and a couple thousand kilometers, so in theory they shouldn't be all that different.


Let's see what TreeMix has to say on the matter. Note that the graphs also include five other ancient genomes: Denisova, Altai Neanderthal, Loschbour, Stuttgart and BR2 (LBA_Hungary).







Admittedly, I'm still learning to use TreeMix. But with that in mind, I'd say the graphs above appear very reasonable, and show outcomes that generally fit with what I've seen elsewhere.

For instance, Denisova harbors something chimp-like that isn't shared with the Altai Neanderthal. This might be a signal of the introgression from an unidentified archaic hominin that has already been reported in scientific literature.

In regards to Kostenki14, the graphs back one of the main conclusions of Seguin-Orlando et al. (ie. the people who first analyzed and published this genome), in that it appears basal to later Europeans. However, the last two graphs suggest that this basal ancestry is not the same thing as the Stuttgart-related Basal Eurasian component described in Lazaridis et al., which, if I understand correctly, is what Seguin-Orlando et al. were saying.

In fact, the basal stuff carried by Kostenki14 seems to be related to the greater part of Ust'-Ishim's genetic makeup. I say the greater part, because Ust'-Ishim also appears to harbor Papuan-like ancestry not shared with Kostenki14.

Is there anything I can do to make these graphs more informative? Perhaps add or take away some samples? Feel free to let me know in the comments below.

By the way, I downloaded the Kostenki14, LBA_Hungary and Ust'-Ishim genomes from Genetic Genealogy Tools. The rest of the samples came from the Reich Lab's Human Origins dataset, available here.

Update 14/11/2014: After looking over the results above and reading the comments below, I made a few changes to the dataset and came up with a couple more graphs that I think are worth sharing. I'm quite certain now that the so called Basal Eurasian ancestry carried by Stuttgart and Kostenki14 can't be lumped into a single component.





See also...

Kostenki14: first genome of an Upper Paleolithic European

Ust'-Ishim belongs to K-M526

220 comments:

1 – 200 of 220   Newer›   Newest»
Maju said...

Taking your TreeMix results at face value, it'd seem that Kostenki is an "Euro-Siberian" ancient mix, what does make some sense. It is possible that much of that early "Euro-Siberian" mix remained up to the Neolithic in Eastern Europe (where the Neolithic seems essentially rooted in the local Epigravettian) and was later spread by the IE Kurgan expansion. Am I making any sense?

What I do not understand well is why Ust-Ishim and hence the "paleo-Siberian" or "ANE" element (correct?) seems so radical relative to the wider Eurasian or macro-population, as if it was some sort of "Basal Eurasian" branch. Any ideas? Do these trees fit well in this aspect with other analyses?

Davidski said...

Ust'-Ishim isn't ANE. The file that I have basically looks ancestral to all Eurasians, including ANE, WHG and ENA.

So it's the real Basal Eurasian, while the Basal Eurasian discussed in Lazaridis et al. is something else, which only makes up part of the basal in Stuttgart, and is in fact somewhat African-like.

I suspect it might be the Eurasian, or semi-Eurasian, stuff carried by the Yoruba and most other Africans.

This is all a lot more complex than what the models we've seen thus far can accommodate IMO.

Onur said...

Thanks for the analysis, David. My suggestion to you would be to add some modern Western Eurasian populations and MA-1 in your next TreeMix analysis.

Chad Rohlfsen said...

I'd say put Dinka in that last one. See if there is a migration from the Dinka to Stuttgart, plus some Ust-Ishim and either Kostenki or Loschbour. Basal Eurasian in Stuttgart may only be those three.

Also, include MA-1, to see if Kostenki is ancestral to him as well.

It does appear that there is something South Asian like about both ancient samples. It would be interesting to see if that Papuan in Ust-Ishim carries over into MA-1, to the exclusion of Loschbour and Kostenki.

Tobus said...

I second the inclusion of MA-1 in the mix.

Seinundzeit said...

This is pretty cool.

I wonder, maybe Ust-Ishim is the closest thing we have to "actual" ASI? Based on formal measures of genetic affinity, he is closest (among all living populations) to the Onge (the primary population used to infer the existence of "ASI"). In addition, ADMIXTURE calculators that have the Onge as a reference population show him to be predominantly Onge/Andamanese. Perhaps the "ASI" construct as understood in Reich et al. was fundamentally confused? I'm beginning to think that ASI isn't really ENA, but just "Basal Eurasian" (not "Basal" as understood in Lazaradis et al., but basal within the larger ANE/WHG/ENA clade, ancestral to all three groups). I recall that Matt speculated something along these lines, in one of his comments on the "ANE Levels across Europe and Asia" piece.

I think Onur's advice is solid, some West Eurasian/South Central Asian populations might make for interesting output. Also, perhaps adding those synthetic ANE samples? I'm not really sure about the latter, but it might be better than using MA1.

As always, your work is amazing. Thank you very much.

Chad Rohlfsen said...

Including an ASI pop wouldn't hurt. We could see if the flow is from or into Ust-Ishim.

Chad Rohlfsen said...

I think that ANE and WHG are going to be a lot more complex than what we think. A UP Near Easterner and a UP South Asian, will change a lot of our thinking.

Maju said...

@Chad: "I'd say put Dinka in that last one. See if there is a migration from the Dinka to Stuttgart"

The "pre-Yoruba" root of the arrow to Stuttgart appears similar to the Dinka-rooted arrow to (I believe) Ötzi in Skoglund & Mälstrom. What is quite intriguing in these trees is that the Africa-rooted arrow does not affect all EEFs (for example pointing to their common root) but only some of them, suggesting some complexity on this matter since very early, what is interesting in itself.

@Sein: "I wonder, maybe Ust-Ishim is the closest thing we have to "actual" ASI?"

ASI and Onge are, for all I know, part of the mainline Eurasian branch. Actually Lazaridis 2014 uses Onge and Dai almost interchangeably for the East Asian reference. So, considering what David said above about Ust-Ishim being a true "Basal Eurasian", there should be no connection. ASI should be either a third branch in the mainline Eurasian sub-tree or a very upstream sub-branch of the West Eurasian branch.

Chad Rohlfsen said...

It does seem like two waves. One might be Dinka like with E, and the other from the zagros, with G. It would be interesting to see if places with a lot of E have more Dinka like ancestry.

Chad Rohlfsen said...

As we get more samples, I think WHG and ANE are going to look alike, with different secondary UP pops, maybe a third for ANE, given its location.

PF said...

Big thanks for all the recent work David.

Can I suggest that you (or someone) establish a standard legend with all the acronyms listed, defined, and briefly explained? I'ma regular reader and it still gets confusing for me, so I imagine it's very confusing for others, especially those just trying to get even a basic grasp on this stuff for the first time. There's just too many acronyms now, and I think a central legend that combines acronyms/populations across papers would be very helpful.

pconroy said...

David,

Fantastic work!

I would like if you added the following to give more perspective on ANE ancestry:
1. Malta Boy - MA-1
2. Clovis-Anzick
3. La Brana
4. Morala-12
5. Kalash
6. Karitaina
7. Lezgin
8. Lambadi
9. Ainu - or Hokkaido Japanese
10. Selkup
11. Udmurt

I would like if you added the following to give more perspective on Basal Eurasian ancestry:
1. Onge
2. Ari Blacksmith
3. Lithuanian
4. Hadza
5. Berber
6. Yemen

Thanks

Chris Davies said...

@Chad: "I'd say put Dinka in that last one. See if there is a migration from the Dinka to Stuttgart"

I looked at the relationship between Kenyan Luo [couldn't find Dinka] and European populations with HLA. I consistently found a strong association between markers found at high frequency in the Luo that reach peak mainland European frequencies in Switzerland and Northern Italy. If I find data from Bavaria I would expect to see a similar pattern.

mickeydodds1 said...

What the Hell is this something 'chimp-like', which is apparently known as an 'introgression' by scientists?
I'm greatly intrigued.

pconroy said...

BTW, the reason I'd like to see the Lithuanian and Hadza in the Basal Eurasian analysis is that if you look here:

http://admixturemap.paintmychromosomes.com/

And take a peek at the Lithuanian sample and then select:
"Full Analysis: Second Event"

You'll see that their model predicts that the Lithuanians are 23% Hadza???!!!

ryukendo kendow said...

This is very nice!

I would like to remind people that Treemix is model-based, and thus where Kostenki roots is acutely sensitive to what populations are included. K14 jumps around quite a bit even with just edge adding.

D and F-stats, which are more consistent, imply that Kostenki could not have rooted with Ust-Ishim, or rooted somewhere close to U-I, or have the major portion of his genome do so, because their relationships wrt modern pops are completely different.

@ Davidski
Thanks. The fourth image is possibly the most beautiful visual demonstration of Basal Eurasian we have.

I agree that Mal'ta should be added.

I somewhat disagree that modern populations, specifically those that we know are admixed, should be added, because that would just introduce so much more to account for in edges over and above the edges we are out to find. If we add MA-1 I think we should drop Kalash, and I think Stuttgart and WHG pops are sufficient for Europeans at this stage.

I would like to say again my opinion that Onge would do more than anything else to clear this tree up.

Chad Rohlfsen said...

Let's try to figure out what EEF, WHG, and ANE are, before going to modern populations.

Chad Rohlfsen said...

David,
Is it possible to do a tree focusing on just Stuttgart?

Some African populations that might work best are the usual, Mbuti Pygmies, with the Yoruba. If you could include the Dinka, Mandenka, and any of the following, it should show something--Sener, Luo, Bioko, Wolof. These people listed have the best links with the HLA. Just to see if it's real or bogus. The Dinka may suffice, if you don't want too many on there.

On the Eurasian side, maybe just Stuttgart, Loschbour, Kostenki, and Ust-Ishim. I am not sure if it is worth including something ASI(Paniya/Pulliyar) or Papuan. I'll leave that up to you. I am not sure how many migration edges it will take, perhaps three or four? Again, it's up to you to play with it. Maybe later, if it's not too messy, Gok2 can go on there, or her own tree. Thanks!!

terryt said...

"Let's try to figure out what EEF, WHG, and ANE are, before going to modern populations".

Agreed. And don't just confine the study to Western Eurasia. South, East and Southeast Eurasia have almost certainly contributed way beyond the regions where the types are now found. And Australian Aborigines must be related to some-one else.

"I think that ANE and WHG are going to be a lot more complex than what we think".

That looks very likely to me too at this stage.

Chad Rohlfsen said...

David,
Do you think KO1 would be beneficial in the Stuttgart tree, along with Loschbour? He is closer to the Near East, and may be closer to Kostenki.

Davidski said...

I updated the post with two new graphs.

Please note, I can't throw too many modern West Eurasians and low coverage genomes into the mix because they have a knack for screwing up the graphs. Also, I don't have the Dinka in this dataset, but I do have the Hadza.

By the way, there's an article on the mystery archaic hominin admixture in the Denisova genome here...

http://planetsave.com/2013/11/22/denisovan-hominins-interbred-with-an-unknown-early-human-species-new-dna-analysis-shows

Davidski said...

Nah, I can't put KO1 into this analysis. Too few markers, and too many of them low coverage.

Chad Rohlfsen said...

Okay. Where are the new graphs? Are those the same four from earlier?

Davidski said...

Scroll down to the 14/11/2014 update and/or refresh your page.

Chad Rohlfsen said...

Got it! Nice! What is the percentage of Hadza, into Stuttgart?

ryukendo kendow said...

I don't think we should add a south Asian population, because they are admixed between ANE, Basal and ENA, too much for this analysis.

@ Davidski
The two new graphs look much better with the D and f stats. However, there are no admixture edges into K14 now.

Can we increase the number of edges? Or does that screw up the explained variance?

Davidski said...

Chad,

If I'm reading this right, then the East African-like edge into Stuttgart in the 3b graph is 15%. In the 5b graph it's 7.3%.

rk,

Pushing up the number of admixture events in these two runs doesn't create edges into Kostenki14. I get all sorts of edges, usually not very informative, but no edge into Kostenki14.

ryukendo kendow said...

@ Davidski
Does pushing up the number of edges result in K14 moving to root with UI sometimes, or is this not the case?

Chad Rohlfsen said...

Nice! That seems more plausible than something Basal to Ust-Ishim. Would removing Lithuanians clean up the West Eurasian clade? Maybe one or two more edges too.

Chad Rohlfsen said...

Is there a way to get an edge from Ust-Ishim and East Africa into Stuttgart, or could EEF just be modeled as something basal to Loschbour with some East African?

Davidski said...

I only managed to get Kostenki14 into a clade with Ust'-Ishim in the one graph I posted above.

It can probably be repeated with different samples, but I'd need to find the right mix each time, and there's an infinite number of options to choose from.

I'll take out the Lithuanians and see what happens.

Gui S said...

If we consider Ancient Middle-Easterners to be a hybrid population of Basal Eurasian (perhaps coming from East Africa, the Sahara or the Arabian Peninsula) and an unindentified divergent Middle-Eastern UP HG population (let's call it AMEHG).
Could it be that K14 is actually more closely related to the putative AMEHG?
It would explain K14's seeming affinity with putative Basal Eurasian (of which we have no actual sample yet), and the lack of migration edge going to both Stuttgart and K14.

A way to test that would be to add Bedouins into treemix and see if there is any migration edge from K14's branch to Bedouins.

ryukendo kendow said...

@ Davidski
If we have no way of producing this edge, perhaps we should do f stats with K14; Loschour, X.

pconroy said...

Can you add Lezgins and/or Kalash?

Chad Rohlfsen said...

It does almost seem more reasonable that the so called "Basal Eurasian" in Stuttgart is nothing more than Ust-Ishim with some East African/SSA, maybe to the tune of 3-10%, so it looks Basal to even Ust-Ishim.

Chad Rohlfsen said...

It was only the lack of a Basal Ust-Ishim and Kostenki that made TreeMix pull 44% between Africans, and the split of East and West Eurasians.

ryukendo kendow said...


@ Gui S
I was thinking the same thing; I suggested a f(Bedouin; any West Asian, K14) and comparing it with other f(Bedouin;any West Asian, X) for X = other modern pops.

However that does not explain why Dai are closer to Loschour than K14.

Davidski said...

Here are seven graphs with no Lithuanians.

https://drive.google.com/file/d/0B9o3EYTdM8lQN0RKZ1gzZjhrRHc/view?usp=sharing

I think running too many edges with two low coverage genomes screws things up bit.

But anyway, I'll see what happens after I add Kalash and Lezgins.

Grey said...

"Is there anything I can do to make these graphs more informative?"

Coming in late but will chip in anyway:

1) something Hadza/San like
2) something Papuan/ASE/Aboriginal like

I think something similar to those two may have been the first and second layers (or at least two of the earliest).

I think there may have been multiple introgressions confusing the issue.

So for example

B) San/Hadza-like (ancestral to)
D) B + archaic 1
E) B + archaic 2

C) Papuan-like derived from B
??) C + archaic 3
or
??) something derived from C + archaic 3

Chad Rohlfsen said...

On number 7, what is the amount of mixture, pre-Ust Ishim, into Stuttgart?

Ju Hoan North, into Kostenki! interesting...

pconroy said...

@Davidski,

Those last graphs - minus the Lithuanians - are very informative, especially when you see that link:
FROM Australia TO Ust'Ishim

Bearing in mind that the Ust'Ishim Y-DNA has actually a match in a man from Telugu (South Indian Dravidian)?!

What that tells me is that, Basal Eurasian may be Veddoid-like?

pconroy said...

@Chad,

Re: Ju Hoan North --> Kostenki!

That's why I suggested Ari Blacksmith earlier, as:
1. I'm Irish (from Ireland) and get 1-2% of this on some admix calculators
2. Ari Blacksmiths are like a caste, in that they haven't admixed with Bantus, so are better representatives of the local East Africans, much like the Druze are in the Middle East

Grey said...

"What that tells me is that, Basal Eurasian may be Veddoid-like?"

One strand at least.

Just a guess but I think the oldest layers might display like they are compressed.

Say the model for the HG phase of all this is that whenever a new layer over-wrote a previous layer the previous layer becomes 20% of the total autosomal then the sequence would go something like:

1st layer (A)
A 100%

2nd layer (B)
A 20%
B 80%

3rd layer (C)
A 4%
B 16%
C 80%

4th layer (D)
A 0.8%
B 3.2%
C 16%
D 80%

The earlier layers would get compressed in the autosomal so what might look like *a* single 4% basal is actually A+B combined.

Davidski said...

Here's the batch with the Kalash and Lezgins (plus Lithuanians).

https://drive.google.com/file/d/0B9o3EYTdM8lQWUxEaW9BU0N4ZHc/view?usp=sharing

This dataset doesn't have the Ari Blacksmiths.

pconroy said...

Bingo!!!

Kalash --> MA1

Exactly like I predicted a few months ago!!!

IMO, the Kalash are not a far flung group of Indo-Europeans or Hunter Gatherers, rather they are the remnant of the original ANE, which left NW India/Pakistan for Central Asia...

Chad Rohlfsen said...

#7 is wild!! Ust Ishim into Loschbour, Loschbour into the root of MA1 and Kostenki, Kalash into MA1.

Seinundzeit said...

David,

What's the percentage for the migration edge from the Kalash to MA1? I wonder why the software is showing a contribution from the Kalash to MA1, rather than a contribution from MA1 to the Kalash?

Thanks in advance.

Chad Rohlfsen said...

number 6 has Papuan in Ust ishim and MA1, to the exclusion of Loschbour.

Chad Rohlfsen said...

That would make sense of the ASE and WHG in MA1, in ANEk7, if it's real.

pconroy said...

What about adding Clovis-Anzick to this?

Krefter said...

This is a bit off subject.

Davidski, do you have any D F-Statistics for early European farmers? I know they share much more drift with mainland Europeans than with people in the east Mediterranean who have more similar ENF-WHG percentages, but is there a way to see if they are overall most related to mainland Europeans.

I'm asking this because I'm wondering if most near eastern ancestry in Europe can be explained by the European farmers or later near eastern admixture.

Chad Rohlfsen said...

I still think Stuttgart basal, whg, and ane will be dropped soon!

Davidski said...

Last batch for today, with BedouinB this time.

https://drive.google.com/file/d/0B9o3EYTdM8lQZnlMMnN2VWNDZ2s/view?usp=sharing

Anzick-1 is in my other dataset, which doesn't have the best MA-1 version. But my guess is it'd just turn out close to a 50/50 mix between East Asians and MA-1.

Krefter,

Early European Neolithic farmers always look more closely related to Europeans than to Near Easterners. There are a number of good reasons for this, like the WHG admixture among the Neolithic farmers, and post-Neolithic ANE and Sub-Saharan admixture in the Near East.

And yes, Near Eastern admixture in most of Europe can be explained by the expansions of Neolithic farmers to Europe, except in southeastern Europe, where there's extra Near Eastern admixture that looks more recent.

Chad Rohlfsen said...

Nice!! Number 7 has Bedouins reaching a little deeper into East Africa than Stuttgart, Kostenki into MA-1, Loschbour into Kostenki... sounds like what Eske/Sikora found!

Davidski said...

Sein,

The edge from the Kalash to MA-1 is 59%. I suspect the Human Origins has too many Kalash that share too much IBD, because I've never seen this outcome using my own dataset with the 6 Kalash that I usually run.

Chad Rohlfsen said...

That one makes MA1 and Kostenki look like your ANE k7.

Chad Rohlfsen said...

Excellent David! Thank you, very much!

As far as that Kalash into MA-1 deal, if the Kalash have a lot of that Ust-Ishim/ASE(~20%), it would explain that. Here is MA-1, on the ANE k7.

ANE 49.08%
ASE 11.22%
WHG-UHG 33.58%
East_Eurasian 1.54%
West_African -
East_African 4.58%
ENF -

Anzick's score made it look like it was mixed with something East of MA-1, in the range of 70%ANE, 7%ASE, 5% ENA, 10%WHG, 8% EA.

I wonder if that ESA, East Africa, and some or all of the East Asian are from the Basal part of Ust-Ishim and Kostenki.

Chad Rohlfsen said...

A Central Siberian genome from 30-40kya, would be a great help.

Chad Rohlfsen said...

Typo:
That is supposed to say ASE, East African, and some or all of the East Asian being reflective of the Basal in Ust-Ishim and Kostenki.

Chad Rohlfsen said...

David,
Maybe for tomorrow, a tree with different populations and Ust-Ishim could find true Basal Eurasian amounts in each group. It might be a reach, but could be worth a shot. Something that hopefully only picks up migration edges from him to current groups.

Chad Rohlfsen said...

David,
One last question for you. On the Bedouin number 7, what is the amount of Basal into Stuttgart and the Bedouins? Is it close to the 44 and 60? I notice that the Stuttgart one doesn't reach back into Africa like the Bedouin one.

Davidski said...

The East African/Basal edge to Stuttgart is 6%, and to the Bedouins it's 26%.

Krefter said...

"The East African/Basal edge to Stuttgart is 6%, and to the Bedouins it's 26%."

Shouldn't basal Eurasian ancestry correlate with relatedness to African then? If basal's relatedness to basal is very very distant, maybe a test with Stuttgart and Loschbour would reveal basal's affinities to Africa.

Chad Rohlfsen said...

Okay. How do you think that translates into SSA? Does that seem plausible? Ust-Ishim, plus that East African, equals the Basal in EEF?

Davidski said...

It seems to me that Basal Eurasian is a very broad term that includes genetic variation on trees like those above from anywhere beyond Ust'-Ishim to East Africans.

I think the vast majority of it in Kostenki14 can be explained by the same ancestry as Ust'-Ishim (so not really Basal in the term we know it right now). On the other hand, most of it carried by Stuttgart comes from beyond Ust'-Ishimm closer to East Africans.

I have no idea how to measure it. Even attempting to describe it gives me a headache.

What we really need is an early Neolithic sample from the Near East to give us a point to focus on, which we can then compare with Ust'-Ishim.

Davidski said...

Actually, no, that probably won't work either. What we need is an ancient genome from Upper Paleolithic southern Levant or the Arabian Peninsula.

Chad Rohlfsen said...

I wonder if it only goes behind Ust-Ishim because of SSA admixture. Not that it's really anything pre-Ust-Ishim basal.

pconroy said...

@Davidski,

"What we need is an ancient genome from Upper Paleolithic southern Levant or the Arabian Peninsula."

I agree!

The best modern substitute would be Druze, Assyrian or Ari Blacksmith

Chad Rohlfsen said...

Would removing the ena pops clear up the branching and edges?

Chad Rohlfsen said...

I wonder if modeling or creating a zombie that is 38% Ust-Ishim and 6% East African, would yield a Stuttgart carbon copy.

Davidski said...

All modern East Africans are very mixed. So are Middle Easterners. Even the Druze have 5-10% of ANE, which I would never have believed if someone told me a few years ago.

We're basically just treading water until more genomes are released, and the higher coverage the better, because the low coverage stuff sometimes gives more problems than solutions.

Ted Kandell said...

David, I think you should add a few samples that are "100%" on certain higher-order components. A "100% North African component" Mozabite a "100% Arabian" Bedouin, and a "100% ASI" Irula to the mix. The North African and "Arabian" components are what's missing for the "Basal Eurasian" admixture for Stuttgart. These would give a much clearer signal than that Hadza, or any East Africans. A 100% ASI (or South Asian Austroasiatic) would localize the "South Asian" component we see in Ust'-Ishim, and possibly separate this out from the shared ancestry with Kostenki.

Michael said...

An ancient genome from anywhere in North Africa would be great. At least there are some mountain ranges, so maybe something usable will turn up someday. We don't know how far reaching this Basal Eurasian group was. The Berber peoples only arrived ~12000 years ago, and Sub-Saharans after that. Maybe the Basal Eurasians were generally Ancient North Africans.

ryukendo kendow said...

@ Davidski
Absolutely amazing.

We have graphs where the admixture edges are very, very stable compared to what we are used to seeing. This probably means that ancient genomes are less admixed, while modern genomes have very complex ancestry; ancient genomes might form a better scaffold on the tree for us to draw stable edges.

Observations:
1. One of the first edges to appear is always from Ju Hoan into Hadza, and Hadza is the closest population in Africa to all Eurasians. This pattern repeats early on in all three runs, and the edge and the rooting for Hadzas are very stable in all subsequent edge additions. This is very interesting, because ethio helix, dienekes and Razib have pointed to the existence of unique Eurasian affinities in the Hadza through ADMIXTURE, and this Eurasian affinity, unlike those of more recently admixed east Africans, is closer to being equally distant from East Asians and West Eurasians, and is not therefore an obviously recent back-migration. I think we can postulate a 'Hadza' substrate, that was the most immediate sister to all OOAs; it might have been the most immediately related people that the OOAs left behind, overprinted by khoisan-like ancestry later. This portion of its autosome might account for why Hadza lies so close to Eurasia despite being a HG population with HG contribution from Khoisan groups.

2. Stuttgart always gets an edge into itself from the Pan-OOA branch splitting from Hadza, and when bedouin is added, the branch goes to both the common ancestor of Bedouin and Stuttgart, and to bedouin again. These edges then become extremely stable to edge additions, and are rooted in the exact same position on all three trees in every run: on the OOA branch immediately after the split of Hadza and OOA. This is probably the proper 'Basal Eurasian' in the Middle East as per Laz.

The demonstration of something so stable and consistent is really elegant and impressive. Congrats David!

3. Whenever present day basal-rich but highly admixed populations like Kalash, Lezgin, or Bedouin are added, they root with the West Eurasian clade, while K14, or both K14 and MA1, are pushed out of the West Eurasian clade. The program then compensates by drawing edges from the remaining West Eurasians to MA-1 and K14. Whenever this pushing out occurs, the rooting of K14, or both MA-1 and K14, are extremely unstable, unlike the previous two cases; they root sometimes with Usty, sometimes to the inside of Usty but outside of both West and East Eurasian, and sometimes inside of the West+East Eurasian clade but outside of West Eurasian itself. Whenever they are pushed out, their drift parameters increase, often markedly for K14. Shifting roots for K14 often occurs in a single run.

This suggests to me that something is happening that is not very tree-like wrt to Ust-Ishim, K14, and present-day populations, unlike in 1 or
2.

ryukendo kendow said...

My proposal for explaining this is that all present-day crown Eurasians all over Eurasia are primarily descended from a rather late wave of colonization from the crown Eurasian root, while earlier genomes we have here are descended from an earlier colonisation from the crown Eurasian root, and are thus all basal wrt to us. Thus the constant tendency to root just outside/around modern (East-Eurasian/West-Eurasian) in an unstable manner. Furthermore, we captured these ancient genomes in the act of losing their original autosomal composition: the later wave of populations that were to give rise to us and which belong very securely to the 'modern/contemporary' East Eurasian-West Eurasian clade in Treemix, introgressed more and more into them, who represent divergent branches basal to the 'modern' clade.

This would explain two things:

1. Ust ishim is slightly closer to ENA than West Eurasian pops, but treemix shows UI as outside both ENA and WEur, instead of being a very basal ENA. Ust Ishim might be an earlier radiation north, who is only just starting to receive ancestry, or left late enough to receive a graze of ancestry, from the first of the modern ENA pops to leave the crown eurasian root after he did, the Oceanians.
2. Stuttgart, bedouin, Kalash, Lezgin, are always securely in the 'modern Eurasian' clade, because the crown Eurasian within them is the recent type, and the Basal is accounted for via edges. On the other hand, Usty and K14 get pushed out of the 'modern' clade. This might be because they carry remnant Crown Eurasian ancestry that is older and basal to the 'modern' clade, rooting nonspecifically just outside the modern (East Eurasian-West Eurasian) clade. This which is their original autosome; now they are experiencing extensive introgression from the first of the 'modern' West Eurasian clades to leave the Crown Eurasian root, after them, WHG. Treemix thus pushes them out of the crown, drawing a line into them from Loschour.

I have a feeling that these early radiations have had most of their genetic material pushed to the margins. K14 was on the way to losing most of it. This might be why most of us are very securely in the 'modern' crown Eurasian clade, and all of us are closer to Loschour than to K14.

Also Chad, I agree that Basal Eurasian in Stuttgart/Bedouin, while real, is probably itself a mixture of many components, possibly paraphyletic. However, these components do not increase affinity to Africans very strongly, indicating that they are basal to crown Eurasians, and are probably not branching from African pops.

Last of all I think this really demonstrates the necessity for us to get genomes wwith simpler ancestry than modern-day ones, or modern pops with as simple ancestry as possible, to generate such good results.

Davidski said...

Both Mozabite Berbers and Bedouins are highly admixed. The reason they create clusters in ADMIXTURE is because they're also highly drifted.

Their admixture easily shows up in TreeMix though.

https://drive.google.com/file/d/0B9o3EYTdM8lQM0pwNzZ2UVk5dGc/view?usp=sharing

All South Asian groups are highly admixed as well, between at least three very divergent groups. They just differ in the proportions of these admixtures and levels of drift. I can't imagine any ethnic groups in India being basal to Ust'-Ishim today.

The only way to figure out the population history of South Asia is to genotype some pre-Neolithic remains from Pakistan and the drier parts of India.

Maju said...

@Mikey: "chimp-like" almost certainly means just more towards the pre-Sapiens root of Humankind, i.e. something like Neanderthal or "Denisovan" (Heidelbergensis). Chimps would be used as outgroup in the analysis, allowing to root the tree.

Maju said...

@David: the new trees are interesting, thanks. It's apparent that there's something African-like in Stuttgart (EEF reference) and that it's likely NE African (i.e. Nubian/Egyptian most likely). The Hadza don't seem a too valid reference (too much of a distinct population).

Otherwise the structure of the trees dances quite a bit in many details, so it's good to remember that they are not "rocket science" and that looking at different trees may help to discern what is more or less certain (traits that repeat) and what is dubious (traits that change).

ryukendo kendow said...

@ Davidski
Would you be running any formal stats soon?

We could make predictions based on these results, and then test them.

Davidski said...

I'm still trying to figure out what would be the most useful stats to run.

I'd really like to see some tangible results, and not the usual stuff that can be interpreted in any number of ways.

ryukendo kendow said...

@ Davidski
Here's something I feel might be unambiguous:

While all Eurasian populations are approx equidistant to Ust Ishim, it is not clear whether present Crown Eurasians are closer to each other than to Ust Ishim. If Ust-Ishim's position in Treemix is correct, this should be easy to verify.

I propose D stats with present day Eurasians between Loschour/La Brana/Mal'ta/Dai, and Ust-Ishim. If all are closer to later pops than Ust-Ishim, then Usty has an analogous position to the basal in K14.

What do you think?

Davidski said...

I can run that in a couple of hours.

What would be the form exactly and which populations would I be testing?

D(Chimp,Mbuti)(Test,Ust'-Ishim)

Or something else?

Chad Rohlfsen said...

I wonder if putting Egyptians on the tree would be informative.

Grey said...

If for the sake of argument the model was

A) tropics
B) out of tropics then out of africa
C) out of S/SE Asia

with the (B) layer submerged by later layers *including* within non-tropical Africa

(pops. in the tropical jungles protected by malaria resistance)

then the (B) layer may be a minority layer within modern Africans as well as Eurasians.

So the closest African population wouldn't necessarily be close to the (B) population but the one with the most remaining (B) signal.

.

Secondly if the (B) layer is hard to spot and also assuming it originally derived from the tropical pop (A) then maybe one of those tropical jungle pops. e.g. Mbuti might have the strongest remaining signal of (B) through being ancestral to it?

(An extreme submersion of this presumed (B) layer including within Africa might be particularly plausible if the (B) population was physically small and the later derived S/SE Asian (C) population was physically much bigger than them.)

(Do Papuans etc have legends of hunting and eating small people?)

.

So maybe (very simplistic but entertaining) something like

(B) = missing link that goes something like: Mbuti-like -> (B) -> San/Hadza-like.

(C) = Papuan/ASE-like derived from (B)

.

"I'd really like to see some tangible results, and not the usual stuff that can be interpreted in any number of ways."

I think there is likely to be a lot of complications added over time but thinking in terms of potentially pinning down and fixing two of the earliest points on the tree I'd say if:

1) there was an early out of Africa pop (B) that was
2) submerged by a derived pop. expanding out of S/SE Asia (C) and
3) the original (B) expansion was mostly coastal then
4) the surviving (B) genes might have their highest frequencies (outside Africa**) among populations in refuge zones along the western coasts furthest from S/SE Asia.

(** or possibly within Africa too)


So the list of western coastal refuge zones (mountain or swamp) with potentially higher frequencies of this presumed (B) signal might include:
north african mountains
coastal iberian mountains
bits of the coastal balkans
brittany/wales
bits of the baltic
ex-malarial zones e.g. Po valley, Sardinia etc

I personally don't think the source pop for this (B) signal still exists but if the model is correct then there should be a common minority signal among Africans and the pops in those refuge zones.

.

Separate to the above was there an east-west split exactly or was it something else i.e. did a broadly similar Eurasian HG population e.g (C) above, emerge out of the preceding genetic soup of waves of expansions and then *diverge* east and west of the Himalayas or was there a similar layer both east and west until the eastern half only got *submerged* by a very successful expansion out of the eastern Himalayas?

Chris Davies said...

@Grey - I think that the West African/North African HLA haplotype in Iberians/Basques/Western France/Brittany/Cornwall/Wales/Western British Isles probably represents a migration <8kya as it is in strong linkage disequilibrium:
http://en.wikipedia.org/wiki/HLA-Cw*16#A29-Cw.2A16-B44

Grey said...

An alternative sequence that might also point at a Mbuti-like pop. might be:

1) tropical African pop. A spreads out of the tropics to the rest of Africa and Arabia losing some of its malaria resistance in the process

2) archaic admixture in Arabia creates pop. B (bigger?) which migrates across Eurasia and also back-migrates into the Africa mostly submerging pop. A outside the jungle zones

3) pop. C derived from pop. B in S/SE Asia expands and mostly submerges pop. B

At this stage the pop. layers might be

Tropical Africa: A
Non-tropical Africa: AB
Eurasia: ABC

(if previously submerged layers follow some consistent pattern where they become a smaller proportion of the autosome e.g. if each previous layer becomes 20% of the combined autosome then at this stage:

Tropical Africa 100% A
Non-tropical 20% A + 80% B
Eurasia 4% A + 16% B + 80% C

etc with other layers over the top later)

In this version
A ~ Mbuti-like
B ~ San-like
C ~ ASE/Papuan/ASI-like

Grey said...

@Chris Davies

"I think that the West African/North African HLA haplotype in Iberians/Basques/Western France/Brittany/Cornwall/Wales/Western British Isles probably represents a migration <8kya as it is in strong linkage disequilibrium:"

Yes if the same pattern - submerged pops. in refuge zones - was repeated with later layers then you'd get a false positive.

I think the actual full sequences might turn out to be something like this (using the second sequence described above and assuming a D expansion out of the Himalayas that only effected east Asia and an E from somewhere in or around Africa)

tropical Africa: A

non-tropical Africa: ABE

eurasia: ABC at first then splitting

east asia: ABCD + whatever

west eurasia: ABC + whatever (missing the D part)

bits of coastal west eurasia: ABC+ E + whatever

so in this model the African HLA you mention could be from either the A or E layer in the sequence.

where
A ~ Mbuti-like
B ~ San-like
C ~ ASE/Papuan/ASI-like
D ~ Tibetan-like
E ~ ?

Sgt said...

LBK-Stuttgart shares more IBD/HIR with Ust'-Ishim than Kostenki 14; don't see this on the graphs -{?}

Davidski said...

It's not possible for Stuttgart to share IBD with Kostenki14 or Ust'-Ishim. They're separated by too many years.

They just share IBS that looks like IBD. But this still doesn't override total genome-wide similarity, most of which doesn't look like IBD.

Helgenes50 said...

They just share IBS that looks like IBD.


It's certainly the same for us when we share with Stuttgart or Loschbour

Davidski said...

There are ways to test for extremely short IBD hits that stretch back 40,000 years or whatever, but you need full genome sequences for that, not genotype data with less than 1 million SNPs.

Chad Rohlfsen said...

David,
How about including Saudis and the Masaai into the Bedouin plot number 7? I think that it is far more likely that there are SSA alleles in the Near Eastern Basal. Probably from mixing, due to proximity. I find the need for another mystery lineage to be unnecessary. Ust-Ishim is Basal, and it probably goes behind him, to represent a time when Eurasians weren't completely separated from Africans. If it was mostly derived from pre-Ust-Ishim, I think we would see more than 6% for Stuttgart.

Chad Rohlfsen said...

I think the Near Eastern UHG was 70% Ust-Ishim and 30% Loschbour like. Or maybe half Ust-Ishim, half Kostenki.

Chad Rohlfsen said...

That small migration edge into Kostenki, from Ju Hoah North, could be a clue, to what separates EEF, from Ust Ishim. That small bit of African.

Matt said...

Davidski: What would be the form exactly and which populations would I be testing?

D(Chimp,Mbuti)(Test,Ust'-Ishim)

Or something else?


I'm uncertain about the form of D stats, that stat would tell us whether Test or Ust-Ishim is closer to the Mbuti group (Ust-Ishim is closer to Mbuti if positive)?
Re: RK's suggestion of "I propose D stats with present day Eurasians between Loschour/La Brana/Mal'ta/Dai, and Ust-Ishim. If all are closer to later pops than Ust-Ishim, then Usty has an analogous position to the basal in K14." would it be running a set of
D(Mbuti, ModernEurasia1)(ModernEurasia2,UstIshim)
stats?

All should be negative if all modern eurasians are closer to one another than Ust Ishim.

I expect everyone should be closer to Ust Ishim if Ust is the ancestor they spread out from... but if Ust Ishim is a sidegroup to the true ancestor they all spread out from, they should be closer together?

Denisovan and "Basal Eurasian" mixture make this difficult (if BE truly exists) but the influence of BE could theoretically be checked using various MHG and ANE (supposed to be BE free).

Btw, I'd also be interested in (Mbuti,Loschbour)(MA-1,K14), (Mbuti,MA-1)(Loschbour,K14) and (Mbuti,K14)(Loschbour,MA-1) if possible (if the K14 sample is in good enough shape for that).

Also, if Yoruban people have Basal Eurasian (or an ancient Middle East component anyroad), then (Mbuti,Yoruba)(Ust-Ishim,Stuttgart) should be positive as an indicator of this?

Matt said...

Actually sorry, re: "if Yoruba people have Basal Eurasian" that maybe would be instead, to be really sure

( (Mbuti,Yoruba)(Ust-Ishim,Stuttgart) - (Mbuti,Yoruba)(Ust-Ishim,Loschbour) )

should be positive (because that isolates the Basal Eurasian part of Stuttgart)?

Still not sure about D-statistics.

Chad Rohlfsen said...

I think that about every SSA should have Basal Eurasian. They haven't found a group with no West Eurasian mixture and traces of Neanderthal. East African SSA should have the most.

Chad Rohlfsen said...

One to One compare

Ust-Ishim and Kostenki
8 49752138 53410953 2.6 726
8 114906174 118287702 2.1 718

Ust-Ishim and Loschbour - Look at all that Basal!!
2 17394049 19509666 1.8 766
2 129255683 132916869 4.9 715
2 177644761 179925980 2.0 750
2 208241313 211914314 2.4 968
3 121633974 124323789 2.0 806
4 32236847 35993946 2.3 777
8 4560250 5479129 2.1 797
8 41705907 53094129 3.7 1464
10 97236674 99261271 2.5 725
10 119354985 121177759 3.6 743
12 20687416 21749447 1.8 753
12 108146362 111810012 2.5 782
13 74695857 76344013 1.7 765
15 61083025 64159531 2.5 705
21 28725753 31130941 1.7 769

Ust-Ishim and La Brana
0!!

Ust-Ishim and Stuttgart
2 177704656 179879954 1.9 702
3 46957741 53856511 2.1 1351
3 73481334 75470851 2.7 702
3 86583234 98390686 1.8 1390
3 121633974 124159213 1.7 726
10 61048410 63380110 2.3 829
11 72005173 74627051 1.7 720
12 20776581 21923154 1.8 771
14 61758158 64686728 2.3 997

Ust-Ishim and MA-1
0!!

Kostenki and Loschbour
0!!

Kostenki and La Brana
0!!

Kostenki and Stuttgart
0!!

Kostenki and MA-1
0!!



It might not be Kostenki with a lot of Basal Eurasian, but perhaps Loschbour has a lot of Basal Eurasian!!

Chad Rohlfsen said...

That could be the clue as to why he is closer to East Asians.

Srkz said...

Chad, Ust-Ishim, Loschbour and Stuttart are genomes with high coverage, Kostenki is intermediate, MA-1 and La Brana are low covered.
Low coverage -> more homosygous snps -> no long chunks.

btw, i don't think these chunks are surely IBS. What is a probability to form a 1000 SNP chunk occasionally, through the random snip combinations? I've made calculations, every 15-20 SNP's probability is divided by 10 (!) So the random 1000 SNP chunk is really impossible.

Maybe the genetic maps are too rough and the 2.3 cM segment is 0,2 cM in reality (cM distances are calculated by interpolation of basic SNPs distances). Or every 1-2 cM segment is a combination of several lesser segments? And we shouldn't forget about pressure selection.

Srkz said...

Extended Ust-Ishim map (added South and South-East Asia) http://s020.radikal.ru/i700/1411/99/39e84d81c245.png

Epicenter is somewhere in Burma. Interestingly, Papuans and Melanesians are close.

Munda-South 57 ---//Confidence: medium
Lahu 56,66 ---//Confidence: low
Malayan 55,72 ---//Confidence: very low
Nivh 55,08 ---//Confidence: very low
Munda-North 54,39 ---//Confidence: low
Hazara 53,34 ---//Confidence: very high
Burmese 53,33 ---//Confidence: high
Xibo 52,71 ---//Confidence: medium
She 52,55 ---//Confidence: medium
Japanese 52,49 ---//Confidence: very high
Cambodian 52,25 ---//Confidence: medium
Dai 52,11 ---//Confidence: medium
Miao 51,47 ---//Confidence: medium
Yi 50,91 ---//Confidence: medium
Melanesian 49,62 ---//Confidence: low
Tu 48,75 ---//Confidence: low
Papuan 47,81 ---//Confidence: high
Hezhen 47,69 ---//Confidence: low
Mongola 47,55 ---//Confidence: medium
Paniya 46,53 ---//Confidence: very low
Han 46,52 ---//Confidence: very high
Naxi 46,28 ---//Confidence: low
Gujarati 46,21 ---//Confidence: very high
Sindhi 44,44 ---//Confidence: very high
Oroqen 44,35 ---//Confidence: medium
Halakipikki 43,81 ---//Confidence: very low
Burusho 43,77 ---//Confidence: very high
Pathan 43,72 ---//Confidence: very high
Daur 43,53 ---//Confidence: medium
Tujia 42,14 ---//Confidence: medium
Sakilli 41,74 ---//Confidence: very low
North-Kannadi 39,77 ---//Confidence: medium
Loschbour 37,86 ---//Confidence: very low
Brahui 37,12 ---//Confidence: very high
Kalash 37,04 ---//Confidence: very high
Makrani 35,13 ---//Confidence: very high
Balochi 34,52 ---//Confidence: very high
BR2 30,11 ---//Confidence: very low
Maya 29,64 ---//Confidence: very high
Karitiana 20,8 ---//Confidence: medium
LBK 20,65 ---//Confidence: very low
MbutiPygmy 3,22 ---//Confidence: very low
BantuKenia 2,64 ---//Confidence: medium
Yoruba 0,98 ---//Confidence: medium
BiakaPygmy 0,3 ---//Confidence: medium
Bashkir 52,65 ---//Confidence: low
Kirgiz 52,25 ---//Confidence: very high
Buryat 51,89 ---//Confidence: high
Yakut 51,07 ---//Confidence: very high
Uygur 50,38 ---//Confidence: medium
Kazah 49,77 ---//Confidence: high
Uzbek 49,66 ---//Confidence: high
Altaian 49,44 ---//Confidence: high
Ket 49,12 ---//Confidence: very low
Selkup 48,43 ---//Confidence: high
Hakas 47,95 ---//Confidence: medium
Mari 47,58 ---//Confidence: high
Evenk 47,15 ---//Confidence: high
Han-North 46,88 ---//Confidence: medium
Dolgan 46,55 ---//Confidence: medium
Even 45,2 ---//Confidence: medium
Mongol 44,67 ---//Confidence: medium
Tuvinian 44,47 ---//Confidence: high
Mansi 43,97 ---//Confidence: medium
Saami 42,11 ---//Confidence: high
Moksha 41,14 ---//Confidence: very low
Tatar-Kazan 41,02 ---//Confidence: high
Tadjik 40,99 ---//Confidence: high
Udmurt 40,9 ---//Confidence: high
Chuvash 40,89 ---//Confidence: high
Nogay 40,14 ---//Confidence: high

Srkz said...

Nganassan 39,89 ---//Confidence: medium
Erzya 39,83 ---//Confidence: medium
Finnish 39,75 ---//Confidence: very low
Estonian 39,58 ---//Confidence: high
Chechen 39,56 ---//Confidence: very high
Veps 39,03 ---//Confidence: medium
German-Austrian 38,72 ---//Confidence: very low
Russian-North 37,89 ---//Confidence: very high
Bulgarian 37,56 ---//Confidence: high
Belarusian 37,41 ---//Confidence: high
Komi 37,13 ---//Confidence: high
Ukrainian-East-and-Center 36,03 ---//Confidence: medium
Norwegian 35,63 ---//Confidence: high
Kumyk 35,48 ---//Confidence: high
Italian 35,46 ---//Confidence: very high
Ossetian 35,44 ---//Confidence: high
Karelian 35,35 ---//Confidence: high
Adygei 35,14 ---//Confidence: high
Balt 35,12 ---//Confidence: high
Abkhazian 34,39 ---//Confidence: very high
French 33,86 ---//Confidence: very high
Russian-South 33,78 ---//Confidence: very high
Lezgin 33,73 ---//Confidence: very high
Azerbaijani 33,73 ---//Confidence: high
Tatar_Lithuanian 33,4 ---//Confidence: very low
Russian-West 33,3 ---//Confidence: medium
Ashkenazi 33,27 ---//Confidence: high
Tatar-Crimean 33,14 ---//Confidence: low
Shor 32,46 ---//Confidence: very low
Swedish 32,25 ---//Confidence: high
Hungarian 32,18 ---//Confidence: very high
Polish 32 ---//Confidence: high
Croatian 31,83 ---//Confidence: very high
Balkarian 31,72 ---//Confidence: very high
Iranian 31,6 ---//Confidence: very high
Greek 30,82 ---//Confidence: very high
Romanian 30,66 ---//Confidence: medium
Basque 30,48 ---//Confidence: very high
British 30,28 ---//Confidence: very high
Georgian 28,38 ---//Confidence: high
Armenian 28,36 ---//Confidence: high
Sephard 26,59 ---//Confidence: high
Sardinian 26,55 ---//Confidence: very high
Ukrainian-West-and-Center 26,08 ---//Confidence: medium
Greek_Azov 22,62 ---//Confidence: low

Srkz said...

Please note: no segments with African populations:
MbutiPygmy 3,22 ---//Confidence: very low
BantuKenia 2,64 ---//Confidence: medium
Yoruba 0,98 ---//Confidence: medium
BiakaPygmy 0,3 ---//Confidence: medium

Chad Rohlfsen said...

Srkz,
What does Stuttgart get?

Chad Rohlfsen said...

I wonder if La Brana would score higher than Loschbour, if it was high coverage.

Chad Rohlfsen said...

Maybe, maybe not. I do remember that La Brana was a little closer to Ust-Ishim than MA-1, according to the D stats.

Are you able to perform d-stats with Loschbour, to see if it scores better than La Brana?

pconroy said...

Sergey,

How does Kostenki-14 look??

Srkz said...

Chad "Are you able to perform d-stats with Loschbour, to see if it scores better than La Brana?"
Davidski is doing a great work, performing D-stats, F-stats, Admixture analysis, PCA-plot etc. I'm just doing some IBD-analysis because no one wants to do it instead of me :DDD

Paul "How does Kostenki-14 look??"
Standard filtering (150 SNPs, >3 cM sums) produce crap due to low coverage. After lowering to 50 SNPs and >2 cM sums Kostenki-14 looks European and Central Asian (still no African connections)
http://s013.radikal.ru/i325/1411/80/9c75759aae62.png

ryukendo kendow said...

@ Davidski
1. The format for the D stats would be D(Mbuti, X)(Ancient, Ust-Ishim).

I propose Ancient to include:
(Brana, Loschour, Stuttgart, Mal'ta, Anzick, Ajvide)

I propose X to include:
(Mbuti, Yoruba, Dinka, Hadza
Lithuanian, French, Tuscan, Sardinian
Lezgin, Georgian, Bedouin, Yemenite Jews
Sakilli, Uttar_Pradesh, Brahui, Kalash,
Karitiana, Ket, Mongola, Dai
Papuan, Australian)

These are populations picked from your THREEPOP dataset according to geography and how admixed they are vs how much of a contributing-to-others population they are.
Whenever a pop from the Set Ancient is being tested, all other members of Ancient are dumped into X and tested as X as well.

The more negative the stat is, the further away they are from Ust-Ishim and closer to later ancient Crown Eurasians. This would help us clarify the position of U-I vs all crown Eurasians.

2. I would also like to propose:
D(Mbuti, X)(Modern, Ust-Ishim).

I propose Modern to include:
(Bedouin, Sardinian, Lithuanian, Kalash, Sakilli, Dai, Papuan, Karitiana). X is defined in 1).

The more negative the stat is, the closer X is to modern crown Eurasians. This would help us clarify the position of U-I vs all crown Eurasians.

3. Also,
D(Mbuti, Kostenki)(Loschour, Mal'ta)
D(Mbuti, Kostenki)(La Brana, Mal'ta)

We already know that Mal'ta and Loschour are closer to each other than either is to Kostenki. That plus these stats above, would clarify the position of Kostenki on the WHG-ANE clade.

If 1) K14 is equi between ANE and WHG, and ANE+WHG are closer to each other than either is to K14, then K14 split off before WHG+ANE split off before each other, aka it is a basal representative of the root population of WHG-ANE. If 2) K14 is closer to WHG than ANE, and ANE and WHG are closer to each other than either is to K14, then K14 is WHG, but with ancestry from outside of the WHG-ANE clade. Since we also know that ENA are closer to Mal'ta than to K14, this basal must be outside of all 'modern' crown Eurasians.

Because Crown Eurasians are almost equally related to Suttgart and K14, this implies either i) Stuttgart and K14 carry the same type of Basal, and K14 has a lot of Basal, (which I believe is unlikely); or ii) K14 has a fraction composed of a population branching before/outside 'modern' crown Eurasian, including the crown Eur WHG in Stuttgart, but after the branching of the Basal in stuttgart, aka 'early-branching Crown Eurasian'.

I suspect this might be the case because, unlike Stuttgart which score purely in Mideast components in ADMIXTURE to represent his Basal, K14's basal is represented by scoring in all pops in Eurasia. All this is discounting genome quality.

Note to Chad and Srkz, I suspect that K14 scoring in Mbuti components in ADMIXTURE and the edge from Ju Hoan to K14, are attempts to represent 'excess' variation outside modern human var introduced by increased Neanderthal.

4. Lastly,
D(Mbuti, ENA)(Papuan, Ust-Ishim)
D(Mbuti, ENA)(Dai, Ust-Ishim)

Where ENA is:
(Karitiana, Ket, Mongola, Dai, Papuan, Australian, Sakilli, Chamar).

This would help us position U-I in an analogous manner.

Srkz said...

Stuttgart (150/3) http://s52.radikal.ru/i135/1411/ac/cd8bfb0a4137.png
Loschbour (150/3)
http://s017.radikal.ru/i439/1411/3d/fee7185881a3.png
Malta (50/2 - low quality)
http://s54.radikal.ru/i143/1411/cd/52060053b391.png

Fanty said...

Hm. Why are some countries missing?
There must be data from those?

Davidski said...

Here are those D-stats. I used Motala12 instead of Ajvide58, because it offers more markers. And I don't have the Dinka.

https://docs.google.com/spreadsheets/d/1_7oby1TCgWlvIC7my9Sy0iyHZx88xDXgvfkGedN2FEg/edit?usp=sharing

Couldn't run the Chamar, Ket and Sakilli in this one, because they're not in the Human Origins dataset.

https://docs.google.com/spreadsheets/d/1pFLsa1p5fIivtjC7q13o04RxDyazcOIg_ax7g70wb_8/edit?usp=sharing

https://docs.google.com/spreadsheets/d/1RinWCPTHV77HCNYNMPFqkBLnYYfVbOKfXnP-d0LjBnE/edit?usp=sharing

https://docs.google.com/spreadsheets/d/14eT00CB3J9sWkkURJvsvc20CONkWsHc84veOhT3i26M/edit?usp=sharing

Chad Rohlfsen said...

David,
Is there any chance you could do a PCA of just ancient samples, including Ust-ishim and Kostenki, with as many hunters and farmers as possible?

Grey said...

"I think that about every SSA should have Basal Eurasian."

I'd have thought so although it may have been partially submerged by the Bantu expansion.

.

A couple of things pointed at in my simplified model.

If the sequence is

A) OoT (out of tropics) into non-tropical Africa and the African border zone

B) derived A + archaic introgression in the African border zone somewhere then expanding and back-migrating

C) derived from B somewhere in S/SE Asia and back-migrating over the top of AB

and populations can be defined as the collection of population layers they contain like sedimentary rocks e.g. A, AB, ABC etc

then in the era just after C has expanded you'd have the following population blocs:

A - tropical Africa
AB - non-tropical Africa and the African border zone
ABC - everywhere else

Which begs the question what comprises the African border zone?

Maximally I'd say Arabia, Middle East, Near East, and a lot of Europe.

Minimally I'd say the coastal zones of the above regions.

So if this model for the very early part of the story is correct the position of this border matters as there would likely be three zones in Europe: the AB zone around Africa, the ABC zone coming up from SE Asia and an intermediary zone between the two so if the model is (roughly) correct then the ancient genomes from this era ought to reasonably fit as either AB (missing south asian / oceanian etc) or ABC or as a mix between the two.

.

Secondly if it's correct that later population layers compress earlier ones I wonder if that effects how they appear to relate.

For example say the rule is that when a population is submerged by another that the previous layer become 20% of the combined autosome then:

AB would be
- 20% A
- 80% B

and ABC would be
- 4% A
- 16% C
- 80% C

so I wonder if software comparing the two populations might calculate the biggest signal of their relatedness i.e. component B and ignore the smaller component A?

Matt said...

Chad Rohlfson I think that about every SSA should have Basal Eurasian. They haven't found a group with no West Eurasian mixture and traces of Neanderthal. East African SSA should have the most.

AFAIK you could test it with D(Chimp,Mbuti)(Ust-Ishim,Stuttgart) and D(Chimp,Yoruba)(Ust-Ishim,Stuttgart).

Significantly positive and these African groups should have Basal Eurasian. Negative and they don't - they could still have Eurasian, just of a sort that was pre-BEu and Ust-Ishim divergence (although is that even a sensible concept?).

And just to be certain compare D(Chimp,Mbuti)(Dai,Stuttgart) and D(Chimp,Yoruba)(Dai,Stuttgart), and D(Chimp,Mbuti)(Loschbour,Stuttgart) and D(Chimp,Yoruba)(Loschbour,Stuttgart)

Chris Davies said...

The source(s) of 'Basal Eurasian' probably originated in a green Sahara (and perhaps also a green Arabia)and over time have dispersed on multiple occasions each time desiccation began.

Davidski said...

Matt, here are those D-stats.

https://docs.google.com/spreadsheets/d/1a7zhpssgK9MEU6-Pf0LYExp_d9dPtv2lFHEOZjtqGjU/edit?usp=sharing

Krefter said...

I used the Hirisplex(older version) on several stone age Euros. Gok2 is predicted to have blonde hair(71.2). Brown hair is the most popular color for EEFs. Also, 4/11 had light eyes, and only 2/11 have A/A in rs12913832.

Central-north EEFs are the earliest evidence of the change towards the modern norm.

http://www.anthrogenica.com/showthread.php?3461-Hirisplex-Hair-Eye-color-predictions-for-stone-age-Europeans&p=58715#post58715

If you wish to comment, don't do it on this page, do it on the link.

ryukendo kendow said...

@ Davidski
Thank you! This is wonderful, a treasure trove of information.

It seems like the Treemix scenario (Ust Ishim ((ENA)(WHG-ANE))) and of Papuan---->Ust-Ishim, is confirmed!

From 1,
i) All modern crown Eurasian pops are closer to ancient western crown Eurasians (La Brana, Loshour, MA-1, Ajvide, Stuttgart NOT in this category) than to Ust-Ishim.
ii) Except for Australo-Melanesians, who are equidistant to ancient crown Eurasians and Ust-Ishim, or in fact slightly closer to Ust-Ishim.
iii) In the Ust-Ishim/aWHG comparison, ENA other than Oceanians, aka Dai, Mongola etc. are, like the rest of Eurasians, closer to aWHG, but they are closer to Usty than Europeans are, and less close to Usty than Papuans are. Pure ENA, e.g. Dai and Mongola, are closer to Usty than ANE-ENA combinations, e.g. Karitiana and Ket.
iv) Mongola, Dai, Papuan and Australian are closer to Ust-Ishim than Stuttgart, while everyone else with ANE or WHG ancestry is closer to Stuttgart.

From 2,
i) Papuan and Australian are closer to Dai and Karitiana and each other than to Ust Ishim. But Papuan+Australian are closer to Ust Ishim than to all West Eurasian modern crown Eurasians with Basal Ancestry, which is all West Eurasians basically.
ii) Mongola and Dai are closer to all Crown Eurasians than to Ust Ishim, except for Sardinian/UI comparison and Bedouin/UI comparison, where they are closer to Ust Ishim than to Sard and Bed.
Karitiana show a similar pattern as Mongola and Dai: closer to Crown Eurasian than to Ust-Ishim. For Karitiana the exception is Bedouin only.
iii) Kalash and Brahui are closer to crown Eurasian without exception.
iv) This is the weird part: Lithuanian, Georgian, French and Sardinian, and Bedouin are closer to all Crown Eurasian than to U-I. However for Papuan/UI comparison, they are closer to U-I than to Papuans(!).

From 4,
All ENA pops are closer to each other than to Ust-Ishim.

ryukendo kendow said...


My conclusions:
A) All Crown Eurasian Populations are closer to each other than to Ust-Ishim(1i). Therefore, Ust-Ishim is likely rooting outside of 'modern' (WHG-ANE-ENA), and the tree is (U-I(WHG-ANE-ENA)). The exception is Oceanians, who are equidistant from ancient Western Crown Eurasian and Ust-Ishim (1ii). So we have to figure out what is happening to Oceanians.
B) Oceanians are closer to modern day ENA Crown Eurasian than to Ust-Ishim (2i), but are shifted towards Usti-Ishim compared to other crown Eurasians. Thus the gene flow is either Oceanian ---> Ust-Ishim, or Ust-Ishim----> Oceanian.
C) We know from f3 stats in the paper that Ust-Ishim is not equidistant from La Brana and Dai/Onge, but are slightly closer to ENA populations, despite his position being outside of (WHG-ANE)(ENA), which should lead us to expect equidistance. While this is not D-stats, (We would need to do D(Mbuti, Ust-Ishim)(La Brana, Dai) for that), this deviation from equidistance tells us that the gene flow is likely Papuan------> Ust-Ishim, not the other way round.
D) This is further supported by 2iii). If Oceanians contributed to Usty, then Usty will gain a generalized affinity to Oceanian's relatives in ENA. If Usty contributed to Oceanians, Usty will not have such a generalized similarity. We know that this ENA scoring closer to Usty than Europeans in Usty/aWHG comparison is due to affinity of Usty with ENA and not Basal pulling Europeans away, because pure ENA in Dai and mongola is closer to Usty than admixed ENA in Karitiana and Ket, and Karitiana and Ket have no Basal.

To reiterate, I think the D-stats strongly demonstrate that UI branched before modern (WHG-ANE)(ENA) branched, and U-I's affinity to ENA in Crown Eurasian is due to contribution by proto-Oceanians into U-I.

Other things of note:
2iv) further makes makes me believe something is going on with Papuans.

Also, the one comparison we have not done, is D(Mbuti, Ust Ishim)(ENA, Loschour). if the above conclusion is correct, then Ust Ishim will be closer to all ENAs than Loschour, and UI would be equally close to all ENA, except for Papuans, where UI will be exceptionally close.

Last of all, I strongly suspect that the 'Basal' in K14 is also like Usty, aka branching before (WHG-ANE)(ENA).

Chad Rohlfsen said...

Who is closer to Ust-Ishim, Loschbour or La Brana?

Davidski said...

Loschbour seems to be a bit closer, as far as I can tell. But keep in mind that La Brana-1 is something like 3.4x coverage, while Loschbour 22x.

MbutiPygmy Ust'-Ishim : La_Brana-1 Loschbour 0.0059 0.872

Here are some more D-stats.

https://docs.google.com/spreadsheets/d/1NNb5kffkTNCsbmmf19voA3YV0JavABVQOuQIY_IZItk/edit?usp=sharing

Chad Rohlfsen said...

Loschbour is geographically closer to Kostenki, so it's quite possible that he has more of the basal type, no matter what coverage.

Chad Rohlfsen said...

I know that KO1 is quite low coverage, but is there a chance that you could check him too?

Davidski said...

Loschbour looks much closer than KO1, but I don't know how reliable this is.

MbutiPygmy Ust'-Ishim : KO1 Loschbour 0.0291 4.457

Chad Rohlfsen said...

Wow! Any chance that you'll do a PCA will all of these ancient samples? No modern pops though.

Davidski said...

I've already tried. The most distinct samples are always those with lower coverage, and apart from that the results don't make much sense.

Chad Rohlfsen said...

Maybe, Northern Europe preserved a hair more of the Basal ancestry.

Chad Rohlfsen said...

That's too bad.

Chad Rohlfsen said...

BTW, I spoke to Eske about the Red Lady of Paviland. I think that he may be checking into testing out that guy. If you don't know who he is, it's a man from Britain, 33kya.

Davidski said...

So the Red Lady of Paviland was a guy?

Chad Rohlfsen said...

Yeah, it was assumed to be a woman early on, and the name stuck.

Matt said...

@ Matt, here are those D-stats.
https://docs.google.com/spreadsheets/d/1a7zhpssgK9MEU6-Pf0LYExp_d9dPtv2lFHEOZjtqGjU/edit?usp=sharing


Davidski, thanks.

With those stats (Mbuti,African)([Ust Ishim,Loschbour,Dai],Stuttgart)

- those statistics are positive
- and always more positive for the same statistic where Mbuti is replaced by Yoruba)
- but they are positive to a very low degree, particularly for the comparisons involving Dai and Loschbour, and with a very low Z score

So the fact the stats are positive supports that Basal Eurasian may have some increased relatedness to Africans relative to other Eurasian clades.

But they are very weak stats.

It looks to me more like this would fit a model where Africans are closer to equally related to BE and other Eurasians, and if there were a contribution to Africans, then the group split off very weakly on the BE side at around the time of BE divergence from other Eurasians (or by the same token, if a group of Africans mixed into BE, pretty strong population structure in Africa).

Another point would be the statistics are more positive for (Ust-Ishim, Stuttgart) than (Dai, Stuttgart) or (Loschbour, Stuttgart) comparisons, which is odd since Ust-Ishim as theoretically unadmixed (archaic or anything else) Crown Eurasian (to use RK's term) should not be different to Dai or Loschbour. Not sure why that could be. But then again the statistic still isn't high, so I don't know if that's interesting.

RK: the D-stats strongly demonstrate that UI branched before modern (WHG-ANE)(ENA) branched, and U-I's affinity to ENA in Crown Eurasian is due to contribution by proto-Oceanians into U-I.

That's all sensible, except I think there is still a question mark over exchange between Ust-Ishim and a relative of Papaun as, it seems hard to remove the influence of admixture into Papuans from a relative of a lineage that led to Denisovans, which would push the Papuans into a more basal position where it could end up weakly more of an outgroup to (ENA(ANE-WHG)) than Ust-Ishim. Not sure if there's a stat that could incorporate that mixture.

If we have a contribution of proto-Oceanians to U-I, then shouldn't (Mbuti,Ust-Ishim)(Papuan,(Modern/Ancient)) for a variety of populations including Dai, Loschbour, etc. tend to be negative with a clear signal (e.g. Ust-Ishim more related to Papuans)?

Chad Rohlfsen said...

I'd use Stuttgart to find basal in SSA. Ust-Ishim could have Papuan or ENA admixture.

Grey said...

"I've already tried. The most distinct samples are always those with lower coverage, and apart from that the results don't make much sense."

Some random thoughts which may or may not be relevant.

Is "basal"

A) the OoA population,

AB) the A + neanderthal mixed population from arabia/middle east

or

ABC) a population (ASE?) derived from a segment of the AB population in S/SE Asia?

If "basal" is basal to the east-west Eurasian split then "basal" would be the latest common layer of those three i.e. the ABC population from S/SE Asia spreading across Eurasia and submerging the earlier AB population layer as they went.

However as both A and AB populations would share some ancestral layers with the ABC basal then any holdout populations around the African border zone might register as basal also.

What might unconfuse the issue would be if some of the populations that register as basal didn't have the third component (assumed to be ASI/Papuan-like) and some did.

The ones that didn't have the Papuan would then (according to this model) be descended from the archaic admixed AB populations who had spread from the middle east and the ones that did have the Papuan would be descended from the ABC populations who back-migrated from S/SE Asia.

.

Surviving (key word surviving) ydna Haplogroups represent a competitive advantage that existed at a certain time and place in the past.

So the AB vs ABC border would likely be pushed back towards Africa over time as the ABC population benefited from their advantage while they could.

.

If the east-west Eurasian split was caused by a 4th population layer D expanding outwards from the eastern Himalayas and submerging the core of east Asia (but less so the coastal fringe) then that would create an ABCD population in core East Asia splitting the east and west (and at the same time cutting off the ABC populations in coastal E/SE Asia from the rest of the ABC layer in south Asia and up the west side of the Himalayas to north Eurasia.

So the same presumed D expansion would cause two splits: West Eurasia vs core East Eurasia vs coastal East Eurasia.

.

Also, pure guess but I wonder if there wasn't some other archaic introgression in the north of West Eurasia separate to the middle eastern neanderthal one and the denisovan Himalaya one - somewhere in the Caucasus or Scandinavia?

terryt said...

"I wonder if there wasn't some other archaic introgression in the north of West Eurasia separate to the middle eastern neanderthal one and the denisovan Himalaya one - somewhere in the Caucasus or Scandinavia?"

I seem to remember the paper that first mentioned the 'Denisova' admixture also mentioned the likelihood of another Eurasian population that must have contributed genes to the modern human gene pool.

"ABC) a population (ASE?) derived from a segment of the AB population in S/SE Asia?"

I think it very unlikely that a population carrying either Y-DNA A or B ever reached as far as East/SE Asia. If it had done surely we would expect to find at least some surviving Y-DNA of either haplogroup.

"Ust-Ishim as theoretically unadmixed (archaic or anything else) Crown Eurasian"

'Unadmixed'? I don't think so. His Y-DNA is SE Asian in origin, or at least South Asian yet his a-DNA is hardly 'pure' South Asian. As demonstrated by:

" I think there is still a question mark over exchange between Ust-Ishim and a relative of Papaun as, it seems hard to remove the influence of admixture into Papuans from a relative of a lineage that led to Denisovans, which would push the Papuans into a more basal position where it could end up weakly more of an outgroup to (ENA(ANE-WHG)) than Ust-Ishim".

If the movement from some 'Papuan' population occurred before the Denisova admixture your problem disappears. Such a scenario is likely because Ust'-Ishim's Y-DNA is closer to that of NO than to other SE Asian K haplogroups, and NO need not have come from SE Asia proper but branched off before its 'brothers' moved south into that region.

Chad Rohlfsen said...

http://dienekes.blogspot.ca/2013/07/west-eurasian-admixture-in-khoe-san-via.html

http://dienekes.blogspot.ca/2014/02/west-eurasian-ancestry-in-eastern-and.html

ryukendo kendow said...

@ Davidski
Thanks! Sure enough, Ust-Ishim is closer to Dai than to Loschour or La-Brana...

But the rest of those D stats in D stats-6 look extremely weird... the stats are 10X more negative for Ami than Atayal? Koreans behaving like Papuans+Australians, but not Han?

I suspect something's wrong here. Could you rerun the stats?

Now could you also add Japanese, Mongola, Ket, Even, and Karitiana?

ryukendo kendow said...

@ Matt
Denisova Admixture is insufficient to make treemix place papuan outside of ANE-ENA-WHG, and it generally never does this.

From the D-stats in 1, 2, 4, when Eurasians are compared compared in an Eurasian/Ust-Ishim comparison, all Eurasians are closer to each other than to Ust-Ishim, except for Papuan/Australain, who are close to equidistant.



Davidski said...

I evened out the numbers of samples and tried a couple of different sets of markers for some of the groups. The results look similar to the last effort, so I'm not sure what's going on.

https://docs.google.com/spreadsheets/d/1NNb5kffkTNCsbmmf19voA3YV0JavABVQOuQIY_IZItk/edit?usp=sharing

I'm not really familiar with D-stats yet, but as far as I know, results like 0.005 don't deviate significantly from zero, so in other words, none of these results are all that far apart.

Anyway, here's what a PCA of the Ami, Atayal, Loschbour and Ust'-Ishim looks like.

https://drive.google.com/file/d/0B9o3EYTdM8lQU1FhYUZBTGxPNUk/view?usp=sharing

ryukendo kendow said...

@ Davidski
Huh.

Could you also include Han, Mongola, Dai, Japanese, Karitiana and Korean in the pca?

Ami is very significantly shifted compared to atayal, much more so that I would expect on a pca like this... the spaces between Atayal and Loschour in this are intercontinental in size. So this is just to make sure we're not capturing something thats real....

Thanks!

Grey said...

@terryt

"I think it very unlikely that a population carrying either Y-DNA A or B ever reached as far as East/SE Asia. If it had done surely we would expect to find at least some surviving Y-DNA of either haplogroup."

Well I wonder.

If the ydna haplogroups represent a population who had a competitive advantage at some point *in the past* then if A spread first and B represents some group somehow derived from A with a competitive advantage over A then they'd back-migrate over the top of A.

Similarly if C derived from one segment of B and had a competitive advantage over than *at some time* then they'd back-migrate over the top of B.

So in most places you'd expect the earliest layers of ydna to be disproportionately replaced with each new layer even while a substantial chunk of their adna and mdna remains submerged in the layers of the combined population.

The extent to which the ydna is replaced will likely be dependent on factors like group size and population size (so the process may have been more dramatic during the HG phase than in later eras) also in most places A, B and C are only the first few layers so A ydna has to survive the B wave and then the later C wave and then each subsequent wave after that.

So I think you'd expect the earlier ydna layers to survive best furtherest from the source of later expansions and in regions with extreme local conditions they had adapted to which protected them from later back migrations hence A in tropical rain forests and B in extreme deserts furthest from the sources of later expansions.

Having said all that I think you're probably right that in the populations with the fewest layers and lots of surviving C e.g. Australia there ought to be some surviving A or B somewhere among them although like the E1 in North Wales I wouldn't expect to find it except in small clusters in specific remote and defensible spots so it might not have shown up yet especially in modern populations but may show up in ancient ydna like C has in the European samples.

nb mdna could spread peacefully by bride swapping between adjacent groups so I don't think it's as good a marker for possible migratory patterns as ydna.

Grey said...

@terryt

addendum

Forgot to mention the HG phase -> farmer phase which is likely to have been the biggest example of this submerging process with regard to the earliest HG phase ydna haplogroups.

Davidski said...

Updated PCA...

https://drive.google.com/file/d/0B9o3EYTdM8lQVThhQVRJdjNTOE0/view?usp=sharing

Grey said...

@ryu

"But the rest of those D stats in D stats-6 look extremely weird... Koreans behaving like Papuans+Australians, but not Han?"

If there was an ABC layer which expanded over the whole area of E/SE Asia and then a later D layer that came out of the eastern Himalayas radiating outwards towards the coast then you might expect a cline from the source to the coast.

If so the predictions would be a larger submerged ABC layer in the regions around the East Eurasian coastline and larger still in Oceania.

Grey said...

ugh always forget a bit

The Ainu might be an example of the latter i.e. an ABC population submerged by an ABCD population with a larger than average (for Japan) component of surviving ABC.

Grey said...

@ryu

"From the D-stats in 1, 2, 4, when Eurasians are compared compared in an Eurasian/Ust-Ishim comparison, all Eurasians are closer to each other than to Ust-Ishim, except for Papuan/Australain, who are close to equidistant."

If there was an ABC layer that covered most of Eurasia except parts of the African border zone but that layer was mostly submerged in most places** by later layers *except* in Oceania then you might expect to see that.

**particularly by the farmer expansion

German Dziebel said...

@Davidski

"Updated PCA..."

Amerindians are clearly most divergent and UI is closer to both East Asians and ancient Europeans (Loschbour) than to Amerindians. East Asians and ancient Europeans are differentially related to UI, with Loschbour being more Amerindian in its relationship to UI.

Grey said...

@terryt

"I think it very unlikely that a population carrying either Y-DNA A or B ever reached as far as East/SE Asia."

Now you mention it I recall you saying something previously about the earliest waves of humans not crossing a certain geographical line dividing Australia from Eurasia and that only happening with later groups.

Chad Rohlfsen said...

David,
Any chance you could throw Stuttgart in that last PCA? Thanks!

Shaikorth said...

German, according to the shared drift stats of the Human Origins dataset Karitiana and Atayal are actually about equally related to Ust-Ishim and for instance Surui are more related to UI than Han. Closeness to UI seems to correlate with non-Africanness more than anything. Drift of modern populations has hijacked the PCA but that hasn't got much to do with relation to UI.

Balaji said...

Davidski,

Tables S11.1 and S11.2 of the Fu paper show D statistics. Can you reproduce these results? Then you can try other present-day populations.

ryukendo kendow said...

@ Davidski
The ANE and high drift in Karitiana has hijacked the last PCA. Could we try it without Karitiana and Mongola?

Thanks!

Chad Rohlfsen said...

I'd say, just add Stuttgart and bedouins. Just to see how they move off the line to loschbour.

Davidski said...

My D-stats are basically the same as in the Ust'-Ishim paper.

Karitiana French : Ust'-Ishim Yoruba 0.0263 6.531
Han French : Ust'-Ishim Yoruba 0.0249 6.932
La_Brana-1 French : Ust'-Ishim Yoruba 0.0208 4.533
MA-1 French : Ust'-Ishim Yoruba 0.0265 5.319
Karitiana MbutiPygmy : Ust'-Ishim Primate_Chimp 0.3431 77.074
Han MbutiPygmy : Ust'-Ishim Primate_Chimp 0.3424 85.754
La_Brana-1 MbutiPygmy : Ust'-Ishim Primate_Chimp 0.3362 64.805
MA-1 MbutiPygmy : Ust'-Ishim Primate_Chimp 0.3341 61.53

I'll run those PCA later today, and I'll only keep one individual from each ethnic group to mitigate the effects of recent drift.

Krefter said...

Davidski, I sent you an email, asking for you to do a few tests with my DNA. Just do control search 71440, and you'll find me. If it's to much trouble just say so here or in an email.

Davidski said...

I get quite a few requests, but unfortunately I don't have the time anymore.

ryukendo kendow said...

@ Davidski
Thanks!

I suggest you do the sets with Basal Eurasian in Stuttgart and etc. separately from the pure Crown Eurasian ones, so we don't get more dimensions of variation.

This whole ENA D stats weirdness is interesting because its so perplexing. Exciting stuff you're doing. We're looking at things that actually surprise us lol.

Chad Rohlfsen said...

I think that Ust-Ishim, Bedouin, Stuttgart, French, and Loschbour would be good. I'm not sure if we should involve something like Hadza, to gauge the amount of pull. It might be informative.

Shaikorth said...

If someone's going to gauge the SSA affinities in "basal eurasian", it's better to go all the way and use Ju-Hoan and not some West and East Africans with more Eurasian affinities, on a dedicaded test. Otherwise Bedouins will do and won't affect the structure of the PCA like Hazda.

German Dziebel said...

@Shaikorth

"German, according to the shared drift stats of the Human Origins dataset Karitiana and Atayal are actually about equally related to Ust-Ishim and for instance Surui are more related to UI than Han. Closeness to UI seems to correlate with non-Africanness more than anything. Drift of modern populations has hijacked the PCA but that hasn't got much to do with relation to UI."

yes we all know what Skoglund wrote about the effect of drift on PCAs. But let's not take it for granted. They didn't model if the drift has always been high in some populations and then populations such as UI became admixed between 2+ drifting populations. What would this do to a PCA? It would keep a long-adrift population such as Amerindians in its corner, while UI will end up between Europeans and East Asians. Considering that Amerindians are closer to modern Europeans than East Asians and modern Europeans presumably have Basal Eurasian ancestry, Amerindians must have split from the parent population before Eurasians did. This is what all PCAs capture.

Chad Rohlfsen said...

David,
After playing around with these Basal signals, you should look to Loschbour and MA-1.
Something to maybe look for, is that gene flow from the high coverage Saqqaq "into" MA-1, after MA-1 is put into the root of Amerindians.
Maybe, a tree with Loschbour, Ust-Ishim, Kostenki, MA-1, Anzick, and Saqqaq, looking for edges all over.

Chad Rohlfsen said...

http://www.researchgate.net/publication/268333605_Human_paleogenetics_of_Europe__The_known_knowns_and_the_known_unknowns

Very interesting paper here!!

terryt said...

"hence A in tropical rain forests and B in extreme deserts furthest from the sources of later expansions".

The problem with that idea is that no Y-DNA A survives in either the South Asian or Southeast Asian rainforests. I agree that 'farming' haplogroups belonging to O have over-run the earlier Y-DNAs in those regions but have by no means replaced them entirely. I would therefore expect that some small amount of A would survive if it had ever been there. Similarly B is unknown in the arid regions of South Asia, Central Asia and Australia. And no matter how you look at it SE Asia and Australia are very likely to be 'furthest from the sources of later expansions'.

"the HG phase -> farmer phase which is likely to have been the biggest example of this submerging process with regard to the earliest HG phase ydna haplogroups".

Agreed. But that layer completely failed to reach Australia.

"Now you mention it I recall you saying something previously about the earliest waves of humans not crossing a certain geographical line dividing Australia from Eurasia and that only happening with later groups".

Correct. But I doubt that was the only geographic line that slowed expansion. The heavily forested, mountainous region round the SW China/NW India region would have been a significant barrier. I also think the haplogroup distribution indicates the Sinai Desert also was a significant barrier to any migration out of Africa, including that of H. erectus.

ryukendo kendow said...

@Davidski
One last thing; for future pcas, could you tell us how much variation is in each dimension?

Thanks!

Grey said...

@terryt

"The problem with that idea is that no Y-DNA A survives in either the South Asian or Southeast Asian rainforests.
...
And no matter how you look at it SE Asia and Australia are very likely to be 'furthest from the sources of later expansions'."

Well the context here is discussing a possible back migration from SE Asia outwards so those places you mention would be near the epicenter.

The places furthest away from that presumed back migration would be Africa and its border zone on the other side of Eurasia.

.

"I would therefore expect that some small amount of A would survive if it had ever been there. Similarly B is unknown in the arid regions of South Asia, Central Asia and Australia."

Despite the above I'd still tend to agree that if the model is correct then some trace of A or B ought to show up in those regions in ancient dna at least - and maybe it will.

How much ancient dna from those regions has been tested?

.

"But I doubt that was the only geographic line that slowed expansion"

Indeed. And getting past a barrier might even have been the spur that led to an advantageous adaption that led to this presumed ASE population.

Chad Rohlfsen said...

If we want an SSA with probably next to no West Eurasian, the Yoruba may be the best bet. South Africans have more Eurasian than West Africans.

Krefter said...

Davidski, have you ever compared ancient genomes to each other, like Laz did? Stuttgart probably has more shared drift with Motala12, Loschbour, and K01 than with La Brana-1. It'd be interesting to see if she has more shared drift with K01 or Loschbour, of course K01 is the favorite.

Is there an accurate way to see if a certain modern Europeans are genealogically more related to MHG and EEF genomes varying in age, location, and ethnicity?

Balaji said...

Davidski,

It seems that the D statistics show that all non-African populations without Middle Eastern ancestry are equally distant from Ust Ishim. However when a population with Middle Eastern ancestry is compared to one without, Ust Ishim is closer to the population without Middle Eastern ancestry.

From Table S11.1, D(Han,French; Ust-Ishim,Yoruba)=0.0243

D(Han,Sardinian;Ust-Ishim,Yoruba)=0.0294

The D value for Sardinians is greater because the Sardinians have more AME than the French. The corresponding values for Georgians should be even greater.

Can we get such D values for some populations from Europe, the Middle East, the Caucasus and South Asia? This will enable us to see the extent of AME in these populations.

Simon_W said...

@ Krefter
I'm not sure if I got you right, but perhaps you might want to try this:

http://bga101.blogspot.com.au/2014/07/model-yourself-as-mixture-of-ancient.html

Shaikorth said...


Chad, that might be true for many South Africans and certainly East Africans, but I think that Lazaridis' world PCA and ADMIXTURE results, plus shared drift stats, support the idea that Ju-Hoan are much less like Eurasians than Yoruba.

Davidski said...

Here are those PCA...

https://drive.google.com/file/d/0B9o3EYTdM8lQOUZIZzY0MEJhY2M/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQeWR6YzBSdDBCX0k/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQYUhIbHFhZ05kOEU/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQUVNYdC1UVlRfOVE/view?usp=sharing

I'll post the D-stats tomorrow.

Grey said...

@Balaji

"all non-African populations without Middle Eastern ancestry are equally distant from Ust Ishim ... Ust Ishim is closer to the population without Middle Eastern ancestry ... The D value for Sardinians is greater because the Sardinians have more AME than the French."


I wonder if "Middle Eastern ancestry" in this context might mean populations in a ring around the African border zone who were living in environments they had developed specific adaptations or advantages for (cultural or genetic) which protected them from the ASE expansion e.g deserts / malaria / islands etc such that the ASE expansion (if it happened) coming up from the direction of India was diverted around the middle east into north and west Eurasia?

(Madagascar might be interesting in this context.)

German Dziebel said...

@Davidski

"...PCAs..."

Sorry, what part of the discussion do they belong to? What are they supposed to represent?

Tesmos said...

Ryukendo kendow,

Can you can get full acces to this paper and check out the stuff?

http://informahealthcare.com/doi/abs/10.3109/03014460.2014.974668

ryukendo kendow said...

@ Davidski
Really appreciate the transparent layout!

1) Well, there seems to be no way to get the PCAs to corroborate the D-stats...

PCAs with ancient samples are very difficult to interpret in any case.

The plots with Bedouin show that U-I is pure crown Eurasian. If Sikora's justification of the position of K14 ('It isn't drifted enough!') is true, then the position of U-I is also explained there.

Could we do one final set of PCAs, this time without Karitiana or Loschour or Mongola, and with Papuan and Australian?

2) It just struck me, that we can figure out the correlation between ANE-WHG ratios in different pops and shared drift with Kostenki, through plotting. If there is no correlation, then K14 is likely undifferentiated, while if there is a strong correlation, then Kostenki is probably some mix of differentiated components.

The Y-axis will be D(Mbuti, Test)(Loschour, Mal'ta), and the X-axis will be f3 shared drift stats for K14 and Test, where Test are all the west Eurasian and Ancient pops in the paper on which f3 shared drift was performed.

If there is a clear line with negative slope, then K14 correlates strongly with WHG. If there is a clear line with positive slope, then K14 correlates strongly with ANE. If there is a clear line with in between slope, then K14 correlates with a certain ratio of ANE-WHG. If there is a cloud, then K14 probably existed in the complex non treelike period when ANE and WHG did not fully diverge yet.

Just an idea.

Matt said...

@Davidski, couple more comments on stats

1) On the topic of all non-Africans without Middle East admixture being equidistant to What do (MbutiPygmy,Ust Ishim)(Kostenki14,Dai),(MbutiPygmy,Ust Ishim)(Kostenki14,Loschbour) and (MbutiPygmy,Ust Ishim)(Kostenki14,Stuttgart) look like?

(or chimpanzee instead of MbutiPygmy, but since we are only interested in Eurasian ingroups, perhaps a human outgroup would give us a more robust signal...)

It would be interesting if K14 shares less with Han than Loschbour, which we already know from previous work, while being equally related to Ust Ishim as Dai and Loschbour, which we know from previous work Stuttgart isn't (or at least, we know present day Europeans like Sardinians and French aren't).

Since that would be another indicator that it is not a Basal Eurasian outgroup to UI that would be making K14 less related to Dai than Loschbour is, but rather Dai and Loschbour share relatedness independently of shared descent from a UI like ancestor.

2) Re: KO1 and Krefter's comments, looking at the f3 stats which include it, it really doesn't seem very much more similar to Stuttgart than Loschbour is similar to Stuttgart in its f3 stats. They're less in a different direction, but not by very much, which is a bit surprising to me since KO1 looked quite a bit closer to Europe in the PCA you previously posted up.

Looking at it, I'd have said KO1 was less than 1/8 admixed with Stuttgart like populations (who are themselves HG admixed). It looks really minimal.

e.g. this PCA based on only four f3 stats [stuttgart, Gok2, KO1 and Loschbour] -

http://i.imgur.com/BF55lKq.jpg

(similar PCA with some more pops - http://i.imgur.com/XgZ7bNK.jpg)

What are the D-statistics like? Is there statistically significantly more or less sharing with KO1 than Loschbour where there ought to be for farmer admixture (less sharing than with Losch for Dai, etc. more sharing than with Losch for ME).

Maybe some of the difference is a lower effect of drift in KO1 than Loschbour, more than Loschbour being much more HG?

Chad Rohlfsen said...

AFA the shared ancestry with the Han, loschbour is something like 3% ENA in the k7. It may not be quite so bogus.

Krefter said...

"I'm not sure if I got you right, but perhaps you might want to try this:"

I meant comparison with F or D-statistics. Laz had very interesting results. Motala-12 and La Brana-1 are pretty distant, and Stuttgart is closer to Motala-12 and Loschbour, which matches with geography. I'm interested to see how K01 fits in the picture, and also how EEFs compare to each other and with MHGs.

I'd also like see if there's a way to test what EEFs and MHGs modern Euros and middle easterns are closest to genealogically.

Seinundzeit said...

David,

If it isn't too much trouble, could you try a PCA with MA-1, Loschbour, Stuttgart, Ust'-Ishim, any Atayal sample of your choice, any Ami sample of your choice, any Dai sample of your choice, any BedouinB sample of your choice, a Lezgin sample (lez3, if you have this sample in the data-set your'e using for the PCAs), a Chechen sample (ch11, if you have it), an HGDP Pashtun sample (HGDP00244, if you have it), and any Kalash sample of your choice.

This would be greatly appreciated. Thanks in advance.

Chad Rohlfsen said...

It looks like a couple of the PCAs with Bedouins and Stuttgart are replicating the 6% and 28% East African results from TreeMix.

terryt said...

"getting past a barrier might even have been the spur that led to an advantageous adaption that led to this presumed ASE population".

I think that is probably correct. The ability to boat across Wallace's line enabled a back-expansion from SE Asia. But barriers have effectively been 'bottlenecks'. For example it is quite possible that CF was the single surviving Y-DNA haplogroup that originally managed to leave Africa. And K2b2 (P) the only one that crossed back over Wallace's Line, etc.

"I'd still tend to agree that if the model is correct then some trace of A or B ought to show up in those regions in ancient dna at least - and maybe it will".

Quite a bit of work has been done on SE Asian/Papuan haplogroups and their distribution through the islands. No A or B yet.

"Could we do one final set of PCAs, this time without Karitiana or Loschour or Mongola, and with Papuan and Australian?"

I don't think there is any real Australian data yet but certainly Papuan should be possible to include. And the result would certainly be very interesting.

Davidski said...

Here are those D-stats and PCA. The PCA with MA-1 didn't work out. Those are always tricky to set up because MA-1 is low coverage.

https://docs.google.com/spreadsheets/d/1FYlt0gxfDbZWPxuNUdKRfBCklTg8VJ-LifCHf3heXyQ/edit?usp=sharing

https://docs.google.com/spreadsheets/d/1aDcQZnBf84bPwSwzjCrgXeugnMYHQwsQP_PdHsSIx7w/edit?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQMjIwMnktdE8xRDg/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQdGhvYjd6a2h1R1E/view?usp=sharing

https://drive.google.com/file/d/0B9o3EYTdM8lQR2hhNnlpTmRwWkE/view?usp=sharing

And here are the shared drift f3-stats between most of the ancient genomes. It's probably best to ignore all comparisons based on much less than 200K SNPs.

https://docs.google.com/spreadsheets/d/1fj5Qvp8GdNUtd3Nyc1Th5yK3zkvqel7DgUbABnUFCyo/edit?usp=sharing

Later today I'll try and see whether these ancient genomes can be modeled as mixtures of each other.

terryt said...

Sorry. A correction:

"it is quite possible that CF was the single surviving Y-DNA haplogroup that originally managed to leave Africa".

I actually meant, in effect, Y-DNA A1b2 or BT(xB). The haplogroup that gave rise to Y-DNAs C, D, E and F. And further to this comment:

"Could we do one final set of PCAs, this time without Karitiana or Loschour or Mongola, and with Papuan and Australian?"

To me many problems will be solved if we can find which of WHG, ENA and EEF are closer to Papuans and which are closer to Australian Aborigines.

Matt said...

Thanks David.

https://docs.google.com/spreadsheets/d/1aDcQZnBf84bPwSwzjCrgXeugnMYHQwsQP_PdHsSIx7w/edit?usp=sharing seems to confirm that K14 shares no less drift with Ust Ishim that Loschbour or Dai, while Stuttgart does share less drift with Ust Ishim than K14.

Fu's paper evaluates closeness to Ust Ishim via the stat (X,Y; Ust Ishim, Chimp), which I would understand to be comparable but opposite in the D score(?).

So in that context (Sardinian,La Brana; Ust Ishim, Chimp) gives 0.0249 with a D score of 3.734, very comparable to the (Primate_Chimp,Ust Ishim; K14, Stuttgart) of -0.021 with a D score of -2.706.

Tentatively, if a Basal Eurasian / Ancient Middle East entered Stuttgart and pushed Stuttgart's shared drift with Ust Ishim below Dai or Loschbour, this doesn't seem to have happened to K14, despite the fact that K14 shares less drift with Dai than Loschbour and MA1 do.

The f3 stats are cool as well.
Although with these, for the comparisons with KO1 I was most interested in, seems unfortunately the SNP overlap doesn't generally meet that 200,000 SNP threshold that you spoke about.

It seems like there is some suggestion that KO1 is similar in its shared drift with Loschbour as La Brana is, but has higher shared drift with particularly the Bronze and Iron Age Hungarians than Loschbour or La Brana (e.g. much higher f3), and also to a lesser extent with the Gokhem and Hinxton samples, but this could be an artifact of differences in SNP overlap. Particularly IR1, BR1, CO1 and KO1 seem close, while NE1 and BR2 seem a bit more unrelated to them - the late Bronze Age and earliest Neolithic sample may have less of a link to ancient Hungary's hunter gatherers than late Neolithic/Chalcolithic, early Bronze Age and early Iron Age people there? (Although BR2 still seems to have pretty high sharing with the HGs generally.)

Chad Rohlfsen said...

I've sent questions to Reich, concerning this Ust-Ishim and Stuttgart conundrum. Hopefully, he has something to add in.

Balaji said...

Davidski,

Thank you very much for your work. The D-stats do show the expected trend with the values being lowest for East Asians and highest for Middle Easterners and Africans.

I took the values for all populations for which you have D stats and for which Lazaridis has EEF values and plotted D vs EEF. There is correlation between the two as can be seen below.

https://docs.google.com/spreadsheets/d/1mRxpCWk9a8UVzvp1yO2mNBED-CalPmt0dG9UbrrkjOI/edit?usp=sharing

Matt said...

@ Chad, be interested to see if Reich gets back to you with, I'd guess they're still digesting K14 and U-I themselves.

I was thinking of this - http://i.imgur.com/QWArHAl.png - as a possible combination that explains what K14 and Ust Ishim seem to show with the modelling data from Lazaridis...

Essentially
1) K14 can't have admixture from outside the UI-WHG-ANE-ENA clade as it is essentially equally related to UI as WHG ANE and ENA, while Stuttgart is less related to UI than the members of the clade
2) WHG-ANE are closer to ENA than K14 is
3) WHG is closer to K14 than ANE is and
4) K14 is no closer to Stuttgart than MA1 is

None of which can be simply explained by modifying Lazaridis's model to add in U-I prior to ENA-W divergence and adding K14 as a W clade.

But that's a lot of admixture edges and is all very complex, so feels unsatisfactory. It does offer a testable prediction though, that if / when Ancient ME / Basal Eurasian is found, it should be further from ENA than either Stuttgart or K14.

@ Davidski, on the f3 stats again, thanks again, I noticed even beyond the 200,000 SNP limits, some of these ancient f3 stats seem odd or at least counter-intuitive to me when compared with moderns.

Like f3(MbutiPygmy,MA-1,Loschbour) with 261785 SNPs is 0.245368, which is lower than the stat you've given before for f3(MbutiPygmy,MA-1,Palestinian) at 0.246977 and f3(MbutiPygmy,MA-1,She) at 0.261601. I'd have thought it should be higher (Loschbour and MA1 closer together than She and MA1 or Palestinian and MA1).

Similarly f3(MbutiPygmy,Loschbour,MA1) again at with 261785 SNPs is 0.245368, which is lower than the stat you've given before for f3(MbutiPygmy,MA-1,Mongola). It's also lower than f3(MbutiPygmy,Loschbour,Karitiana) at 0.257724 (also true for Pima Indians), which seems strange. MA1 also comes out as sharing less drift in f3(MbutiPygmy,Stuttgart,Target) than Malay, She or Japanese (slightly higher than Papuan or Ust Ishim), with SNP overlap 260054.

I guess perhaps the error levels mean the f3 stats cannot be directly comparable over this distance?

Davidski said...

The Eurogenes dataset based on illumina SNPs gives higher shared drift stats for MA-1 than the Human Origins dataset based on Affymetrix SNPs. The difference is probably due to the lower number of SNPs and higher number of homozygous alleles in the MA-1 sequence from the Eurogenes dataset.

https://docs.google.com/spreadsheets/d/121Ug-u4hEeF2XouaBshD0x-TkqV_pYrXhEDUDO1wujk/edit?usp=sharing

https://docs.google.com/spreadsheets/d/1V5bP7lUjiFa34Rwr2KecuRslA_zDvc5U8PDb0YqSACs/edit?usp=sharing

Matt said...

Thanks. It seems that f3 stats with samples can't be compared very closely without similar levels of SNPs being considered.

Matt said...

Also, David, just for completeness on the comparison with the D-stats5 you ran off for me - would you mind running D(Primate_Chimp,MbutiPygmy)(Ust_Ishim,Dai) and D(Primate_Chimp,MbutiPygmy)(Ust_Ishim,Loschbour)? If you get the time at any point.

Looking at Dstats5 again I'm finding it really curious that Ust Ishim seems less related to Africans than Stuttgart is - the statistic actually looks fairly robust- while this seems much less the case when comparing how related to Africans Dai and Loschbour are relative to Stuttgart.

Davidski said...

Matt,

Primate_Chimp MbutiPygmy : Ust'-Ishim Dai 0.0185 5.398
Primate_Chimp MbutiPygmy : Ust'-Ishim Loschbour 0.0163 3.529

Balaji said...

Davidski,

Primate_Chimp MbutiPygmy : Ust'-Ishim Dai 0.0185 5.398

This is interesting indeed. Mbuti are closer to Dai than they are to Ust-Ishim! Could it be that some population contributed to both Mbuti and Dai some time after Ust-Ishim?

You had earlier given the following statistics.

Chimp Loschbour : Dai MA-1 0.0504 7.792

Would it be possible to calculate D(Chimp,Dai;Loschbour,MA-1)? This will tell if Dai are equally distant from Loschbour and MA-1 or not.

Matt said...

Thanks. Via D(Chimp,Mbuti)(Ust Ishim)(Target) it looks like (I think) MbutiPygmy shares more drift with recent Eurasians compared to Ust Ishim, with somewhat similar excesses in Loschbour (D : 0.163, Z : 3.529), Stuttgart (D : 0.0190, Z : 4.485) and Dai (D : 0.0185, Z : 5.398) relative to Ust Ishim, when chimp is the outgroup. I'm assuming this is a fairly robust / meaningful signal as (although lower) it seems on the order of the signals picked out in the UI paper as showing a greater similarity of UI to Loschbour and Dai than Stuttgart.

That seems quite curious as L, S, and D in theory should descend from populations that separated before U I, then stay separate.

Chris Davies said...

@ Balaji - "This is interesting indeed. Mbuti are closer to Dai than they are to Ust-Ishim! Could it be that some population contributed to both Mbuti and Dai some time after Ust-Ishim?"

I have HLA haplotype data from China Southwest Dai. One prominent HLA haplotype in this population is: A33-B58-C3-DR3-DQ2. This is very different to all other haplotypes in the Dai and clearly originates elsewhere.

The haplotype is in very strong linkage equilibrium in the Dai, with a frequency of 7.7%.

The whole haplotype is also found in Omanis, Emiratis, Iran Baloch, Pakistanis (peak frequency in Baloch, but also found in Burusho, Kalash, Pathan etc.), plus Cantonese, southern Chinese, Native Taiwanese, Thais, Malays, Vietnamese, Koreans, Japanese etc.

All components of the haplotype are found at peak frequency and diversity in Africa. A33-B58-C3 is found in Gambia, Cameroon, northeast Africa, northern Italy, Switzerland, Austria, Croatia, Romania, and Turkey.

Chad Rohlfsen said...

Chris,
That is pretty interesting. The Central Asian Neolithic met the East Asian rice farmers in Western China. Maybe it is the East Eurasians that have Loschbour like admixture from the farmers and some WHG in Northern Hordes, and not East Eurasian into Loschbour.

Chad Rohlfsen said...

It would be interesting to test this on Treemix, using Loschbour, Kostenki, La Brana, with Mongols, Dai, and others.

Balaji said...

Chris Davies,

Thanks for the information on HLA haplotypes. Could these haplotypes be subject to selection?

I, probably like Matt and others, had expected Mbuti to be closer to Stuttgart and equally distant from Ust-Ishim, Dai and Loschbour. Since this is not so, it appears that there was some gene flow over continental-scale distances after 45,000 BP and before 10,000 BP.

Davidski said...

Balaji,

Primate_Chimp Dai : Loschbour MA-1 0.0015 0.245

«Oldest ‹Older   1 – 200 of 220   Newer› Newest»