search this blog

Monday, July 11, 2016

Layers of Ancient North Eurasian-related ancestry in East Asia

Back in May I hypothesized that present-day East Asians were prehistoric hybrids of partly Ancient North Eurasian (ANE) origin. I got the idea from a series of TreeMix runs (see here).

This was essentially confirmed recently in the Lazaridis et al. 2016 preprint. Refer to page 147 in the paper's supplementary information PDF here.

However, based on more recent TreeMix runs featuring data from Lazaridis et al., I'd say the situation is more complex than just some minor ANE-related admixture in East Asians. I suspect now that all East Asians, including even the Onge, an ancient isolate population from the Andaman Islands, harbor significant ANE-related ancestry that may have arrived in East Asia in separate waves.

Here's what I'm talking about. Note that all of the samples on the East Asian node - Upper Paleolithic west Siberian forager Ust-Ishim, Han Chinese and Onge - are influenced by a massive migration edge from the base of the AG3-MA1 or ANE branch. However, as per the second graph, only the ancestors of more northerly East Asians, like those of the Han, appear to have been recipients of the latest ANE-related admixture into East Asia.

Indeed, when I add the Natufians from the Epipaleolithic Levant to the analysis, Ust-Ishim and the East Asians join AG3-MA1 on the same branch, but now receive a 36% migration edge from a point basal to all Eurasians. This is not admixture from the hypothesized Basal Eurasian clade, but probably from another basal clade, specific to East Asians, which I'd say occasionally shows up as pseudo Sub-Saharan admixture in East Asians.

But obviously, we'll need a solid selection of ancient genomes from across space and time in East Asia to confirm these results. Rumor has it that they're on their way.

See also...

East and West Eurasians seperated at least 45,000 years ago, but...


ryukendo kendow said...

Thanks for this analysis, its not often that people perform runs centered on phenomena surrounding East Asians.

For the final tree, do you mind increasing the number of edges to 2 and 3?

Davidski said...

Up to four edges is very interesting; after that it gets weird.

ryukendo kendow said...

Thanks, this is very interesting! If this is the actual pattern, perhaps the major part of ENA, which seems to be ANE if we take these treemix at face value, reached its position via a counterclockwise movement around the Himalayas, for which signs have been accumulating for some time.

Perhaps Ebizur's past comment about how E Asian Y-Haplogroups being even more polyphyletic than W Eurasians, which seem not to fit with the idea of W Eurasians being mixtures and East Asians not, gets some backing here. Very interesting also what's going on with Ust-Ishim and the rest of the Crown Eurasians, an extremely different topology, which still can maintain the similarity of modern Crowns away from U-I. This does seem to suggest that Ust-Ishim should be quite close to Onge though, which we should test w stats...

If you have the time or energy, do you mind adding Mota and a S Asian, say a Paniya or an Indian_South? I think these will be *really* interesting, especially as to how Paniya are resolved and the placement of the E Asian edge with respect to Mota.

ryukendo kendow said...

It almost looks like the 'great coastal migration', actually. Maybe a Papuan would intensify the effect, where the Papuans preserve the one-third non-ANE 'African' ancestry in ENA the best?

bellbeakerblogger said...

I wonder if this is the source of Neanderthal or Denisovian admixture in East Asians. Is there any way to estimate what percentage of ANE is Neanderthal?

Davidski said...


Lots to think about here.

Can't really add Papuans because they have too much Denisovan, so I'd have to add Denisovan as well, and then things get more complicated.


The high coverage Altai Neanderthal doesn't have any Early Modern Human admix. More likely it has something from an early wave of human migrants from Africa who died out. But even that's likely to be minimal.

MA1 has 2-3% Neanderthal, and so does Ust-Ishim, it's just that it's in bigger chunks, and thus more recent, which makes sense. Neither has much Denisovan.

This doesn't mean, however, that East Asians didn't get their relatively high Neanderthal admix from an ANE-related migration wave. ANE-related could mean several things.

aniasi said...

Question: Does this mean that ANE (ancestral north eurasian) isn't really Northern at all?

ryukendo kendow said...

@ Davidski

Wow, thanks for these trees! Very nice, no 'stickies', very clean clades and branching, very beautiful and transparent and and stimulating to the mind, if I may say so.

...Hmm, it seems that even though the various trees move quite a bit they're still very synoptic though, representing the same phenomena. It looks like variation in Eurasia is 'Fan-shaped' if that makes any sense, with a core group of closely-related populations in the centre of the 'Fan' and getting more contribution from Basal clades on either side as one heads towards East Asia and the Middle East.

In particular, it looks like there is a Basal on the West Eurasian side, and there is another on the East Asian side, while the centre of the 'Fan' comprises WHG-ANE, which is also now extended to include ~60% of Han, Ust-Ishim and Onge as well, with the rest of Han, Ust Ishim and Onge coming from a Basal II, or East Asian Basal, or Australasian, or whatever you choose to call it.

E.g. in tree no. 2, we already get a (WHG, (ANE, ENA)) fan-centre pattern with 60% contribution from Australasian/'Basal II' to ENA, while on the Middle Eastern Basal side, Israel Natufian is placed in the Basal position in Eurasians, accounting for that; then the Basal attraction between Israel_Natufian and Iran_Neolithic is addressed at tree 3 with edges between these. Tree 4 looks different superficially, with East Asians pushed out, but its really the same thing, because ANE continues to send the ~.5 edge into ENA while ENA is shifted out into the 'Australasian' position, addressing their dual origins in the same way. Israel Natufian gets pushed into West Eurasians, but now has a 40% typically 'Basal Eurasian' edge. Tree 5 resumes the topology of trees 3 and 2, with East Asians pushed back to the centre of the Fan beside ANE, while the Australasian half of ENA addressed by the edge, instead of the ANE half.

So it seems like theres some kind of arc-shape. Where the trees disagree, is on whether the East Asian basal or the West Eurasian Basal splits off earlier: trees 2 and 5 support ENA's basal/'Basal II' being earlier, trees 3 and 4 West Eurasian Basal being earlier. While there is an attraction between Israel Natufian and Iran Neolithic, necessitating at least one Basal clade in West Eurasia, it seems like the Basal in West Eurasia is not a single clade as well, possibly multiple clades, from the way the two are often treated differently.

Hmm, this gives us a great deal more degrees of freedom, lots of things are up in the air. It doesn't seem to contradict stats either, except for one: isn't even Bedouin closer to East Asians than they are to Ust-Ishim? Don't all West Eurasians strongly favour East Asians over Ust-Ishim? If so, it would seem to support Ust Ishim as being an outgroup to the fan centre, as seen in tree 2, instead of being attached to East Asians.

And it looks like time spent on the Mammoth Steppe in the cold led to disproportionate success all over the Eurasian landmass?

ryukendo kendow said...

Seems like the survival of Y haps D in the far Eastern reaches gets some smoking gun; also Rasmussen's model of Papuans splitting earlier than East and West Eurasians split off from each other, but extensive gene flow between Papuans and ancestors of East Asians+Amerinds. Also this seems to be relevant to the very weird bifurcated nature of East Asians w N Chinese clading with Europeans and S Chinese clading with Papuan, in the earliest research all the way back to Cavalli-Sforza, perhaps an artifact of the inability to represent reticulation via trees? This may explain some of Anders Palsen's haplotype-based research as well.

Also it seems like more and more stuff is happening in Siberia, with some West-Eurasian-Central-Asian-Siberian connection, like in Alan's model of archaeological patterns in the Eurasian Paleolithic. Which I strongly doubted before in favour of pincer movement north around the Himalayas from India and SEAsia--which is completely anarchaeological actually--but the Siberian model kinda makes sense now. While the southern tier of Eurasia is a reservoir for various clades basal to a large wave of recent ancestry expanding south from Siberia.

Chad Rohlfsen said...

East Asians do approach significance in distance from Ust Ishim, compared to GoyetQ, but not as much as WHG. I asked Nick once about the possibility of BE in all modern and post UP Eurasians. I'll have to look for his answer. It would also make sense of WHG and all West Eurasians, even those with BE being closer to East Asians than Ust Ishim.

Chad Rohlfsen said...

Those qpAdm results showing WHG as a mix of Anatolian, Aurignacian, and ANE looked as good as modeling Bronze Age Euros as Yamnaya + MN with the same outgroups.

Matt said...

Interesting. So this new data and model in a sense takes us back to the "Basal Eurasian admixture in Onge / ENA models", that were rejected originally in Laz 2013.

What happens when we next add Karitiana to the structure?

RK: If this is the actual pattern, perhaps the major part of ENA, which seems to be ANE if we take these treemix at face value, reached its position via a counterclockwise movement around the Himalayas, for which signs have been accumulating for some time.

Seems like it could've just as easily happened already in Central / South Asia - the whole thing is common to the whole Ust Ishim-Han-Onge clade.

I find it all a little puzzling though. Wouldn't we expect D-stats of the form

D(Mbuti,Han/Onge/Ust Ishim,ANE,WHG)

to be quite significantly negative if that topology in these treemix is true?

The original problem that motivated Laziridis 2013 to generate Basal Eurasian into EEF in the first place was that stat being non-significant...

Does this relate to the choice of Bichon and Villabruna as representatives of WHG, which are less related to ENA? I guess this choice is motivated by Bichon and Villabruna being in theory non-admixed by EHG...

Grey said...

fascinating stuff

Seinundzeit said...


Tremendous stuff.

Like almost every analysis you've tried, I'm sure this is going to be verified with aDNA (from East Eurasia).


That probably explains it, as Villabruna is quite a bit more distant from East Asians than Loschbour, and is also less MA1-shifted than that sample.

With the Onge, much could be solved. For example, MA1 is closer to the Onge than WHG/EHG are to them, so I wouldn't be surprised if they (and populations related to them) have admixture from something really basal within the ANE cluster.

If so, ASI could really be an Onge-like population, but with much more ANE-related ancestry than the Onge. I think this would explain the consistent affinity that keeps appearing between MA1 and ASI-rich South Asians.

jparada said...

I noticed in extended data figure 7 in the Lazaridis 2016 preprint, where ANE admixture in East Eurasians is shown, that MA1 is Onge-shifted wrt AG3, even sitting along the same line East Eurasians do. This makes me suspect gene flow went both ways.

Seinundzeit said...

It's pretty interesting that Han and Indian_South are joined together, rather than Onge and Indian_South.

Rob said...

Having had time to look at it closely, this is fantastic, well done Dave & thanks for your explanations (RK, Matt, Sein).

As Ryu said, this supports the idea that there was a rapid southern dispersal, with an eastern component (common to U-i, Han, south Indian), and a western basal (-> Iran and Levant Neolithic). Then you have this 'Crown Eurasian' branch (if that's what its referred to as), but the important thing to notice is that in each tree it specifically branches off the Iran_Neol spectrum of Basal Western rather than Levant (Israeli Natuf).

This suggests that Crown Eurasian branched off from somewhere in the vicinity of Iran, before itself then splitting further into eastern (Afontova cluster, & northern components of Han, Usty, Onge) & western (K14, WHG). Its hard to put this into a coherent narrative, as this only gives as relative position, not absolute.

This supports models suggested by Otte & Kozlowski that the Aurignacian in Europe arrived from central Asia/ Iran rather than the Near Eastern Emiran - its just a shame the U/P of central Asia is still not too well researched. I think this is also what Ryu was referring to about something he discussed with Alan, and also what many might have guessed at (albeit simplistically) by looking at the Admixture graph of many of the earliest Eurasian genomes.

Other notable aspects are that the 'European' Palaeolithic type admixture seen in Natufians (but absent in Iran Neol.) comes from a branch leading to Villabruna / Bichon, meaning western farmers received northwest Eurasian admixture during the post-LGM ("Epi-Gravettian) period (18- 12 kya), imaginably from a movement in the Balkan - Black Sea region. (with regard to WHG , EHG, ANE interactions, I still think we are missing a vital peace of puzzle from LUP Ukraine, in that something which might subtract from the apparent 60% Af-G admixture in EHG)

Further, ANE seems to be a parallel to the non -basal components of U-I, Onge, ASI, the latter of which all have Basal, but ANE does not.

Now, lets recall that Ust-Ishm (42 ky BP) is found in western Siberia (and has basal), but 20, 000 years later, Mal'ta (ANE) does not, which is along a similar latitude, further East. So it looks like ANE replaced U-I type people at some point during the Upper Palaeolithic of the Siberia - Altai region (42- 22 kya), which is also broadly when UP Europeans arrived to Europe, perhaps slightly later. Again it points to from around Iran, with ANE going north along the Inner Asian mountain corridor and northern East Asian going south of the Hindu Kush.

Finally, in terms of archaic hominids, Denisovans admixed with one of the sub-branches of Basal far eastern (Melanesians & Aborigines); whilst Neanderthal only directly with crown Eurasians (but also indirectly with Melanesians by way of Denisovans (?))

David (or anyone)

* Can you include CHG (although not too much mystery for it)

* was anyone able to test whether Oase had any 'basal' ? to see if any of the very ealiest (pre-CI) 'Europeans' had any

Chad Rohlfsen said...

I'm wondering if they tested this in qpAdm, because I've tried everything for Han from Onge+MA1, Onge+UstIshim+MA1, UstIshim+MA1, and all were badly infeasible.

Chad Rohlfsen said...

However, ENA into ANE and definitely EHG fits very well.

Davidski said...


Adding Karitiana makes the East Asians look unadmixed. Here's the best tree with that topology.


They did test it with qpAdm. The results are in the supp info.

Davidski said...


Tree with CHG included.

ryukendo kendow said...

So two of the trees contradict. Matt's tree is much more orthodox and compatible with the 'classical' single Basal Eurasian scenario that we've had so far, with Iran_N in the Basal position sending edge to Natufian. Interesting also that there is the same very basal W Eurasian population sending branches to Iran_N and South_India, which matches our observation of Iran_N and S Asians having exceptionally low drift with EHG and Ami compared to other Eurasians, while being disproportionately close to U-I for their level of drift with EHG and Ami, leading to the population 'X' in C Asia scenario.

David, for Matt's tree can you increase the no. of edges and see what happens? Alternatively, to clarify things like the authors do in ADMIXTUREGRAPH where they add a population to discriminate between alternative topologies, could we drop Bichon and Biaka and add Denisovan and Papuan?

The residuals for my trees and Matt's would be interesting to see as well. Thanks!

Davidski said...

I won't have time today. I might get back to this later in the week.

Chad Rohlfsen said...

Okay, I see what they did there.


right pops:

best coefficients: 0.920 0.080

std. errors: 0.018 0.018

fixed pat wt dof chisq tail prob
00 0 3 0.979 0.806341 0.920 0.080

ryukendo kendow said...

Thanks David. Chad, If I give a .txt file of the all the African comparisons to you, which I expect will be highly tedious for you to type by hand, would you be willing to run the stats as a list? Just to reduce the workload. I would have to look at a list of the population names for that though.

ryukendo kendow said...

Hmm, this should also probably reduce our assumption that 'symmetries' in D and F4 stats are strong evidence of clean clades as well, now that we know that symmetry between the Loschbour/La Brana plus Ma'lta pair to the East Asians plus Onge pair is really just a coincidence, and once we add Bichon and EHG and Kostenki and GoyetQ we get widespread asymmetry and fluctuation in distance to East Asians and Ust-Ishim among West Eurasian HGs.

Matt, what do you make of the increased Denisovan in S Asians though? Especially now Onge and SE Asians don't have this? Don't you think this rules out a Southern Route for East Asians?

Chad Rohlfsen said...

Adding an extra 0.4% Neandertal for a better fit.


right pops:

best coefficients: 0.915 0.080 0.004

std. errors: 0.019 0.018 0.005

fixed pat wt dof chisq tail prob
000 0 2 0.267 0.875229 0.915 0.080 0.004

Jaydeepsinh Rathod said...


"If so, ASI could really be an Onge-like population, but with much more ANE-related ancestry than the Onge. I think this would explain the consistent affinity that keeps appearing between MA1 and ASI-rich South Asians."

I think a better explanation could be that the MA-1 group migrated into Siberia from somewhere down south (perhaps from SC Asia) as a WE population, and then received admixture from an ENA population, resulting in its affinity towards all ENA groups. An admixture from ANE like population into East Asians is also likely but that cannot alone cannot explain the Onge affinity of MA-1.

"It's pretty interesting that Han and Indian_South are joined together, rather than Onge and Indian_South."

I am actually puzzled by this. Is it possible that the Indian_South has Austro-asiatic groups of India, which have received heavy SE Asian admixture in the last 5000 years ? But that admixture must have come from SE Asian Austro-asiatic groups. Are these Austro-Asiatic groups closer to East Asians than they are to Onge.

One last point, Onge are clearly more closely related to SE Asians than to South Indians, being most closely related to the Negrito populations of Malaysia. Therefore, ideally, without any admixture, South Indians should be an outlier to both Onge & Han.

Davidski said...

Onge look mixed, between something related to MA1 and another population that isn't closely related to MA1.

So if MA1 came from South Asia or has ENA admixture, then why is it missing that basal input that contributed to all East Asians, South Indians and Ust_Ishim?

Jaydeepsinh Rathod said...


"Matt, what do you make of the increased Denisovan in S Asians though? Especially now Onge and SE Asians don't have this? Don't you think this rules out a Southern Route for East Asians?."

I think this has got to do with some very ancient affinity of South Asians to Papuans/Oceanians.

In the admixture graphs of Lazaridis et al 2013 & Haak et al 2015, when the Papuan related component emerges at lower Ks, the only other population group that shows significant Papuan component are the South Asians, including even groups like the Pathans etc.

Even in the recent paper on SE Asians, this Papuan/Australian component is clearly visible in all South Asians.(Supplementary Figure S3-S8) layered_population_structure_in_Island_Southeast_Asians

Also, in the above paper, in figure 1a, we can see that Indians, North as well as South, seem to preserve in trace quantities, all SE Asian components, including the Papuan/Australian, which is absent even in most SE Asian populations. This is me suggests that the Papuan component visible in all South Asians is an artifact dating to the period of shared ancestry between SE Asians, Australians/Papuans & South Asians.

Somehow, South Asians have managed to preserve the signature of this archaic shared ancestry which even SE Asians have not managed to, and this maybe the reason for its relatively elevated Denisovan admixture levels.

Matt said...

@ Davidski, thanks for that treemix. Surprisingly, it does look roughly more like the usual models, with a Basal Eurasian clade contributing to Natufians and Iran_Neolithic; more to Natufians, less to Iran_Neolithic, with Iran_Neolithic getting ancestry from a basal position within the Paleolithic+MesolithicHG-ANE clade.

RK: Matt, what do you make of the increased Denisovan in S Asians though? Especially now Onge and SE Asians don't have this? Don't you think this rules out a Southern Route for East Asians?

Not sure. I think substructure in situ not absorbed by an initial wave wouldn't be too implausible. Increased selection against Denisovan ancestry in East Asia? (which seems less plausible). I can see why it is simpler to assume another route though.

Rob said...


When you get a chance, can you add also Goyet ?

Can't wait for those Mesolithic south Asian results to come through ;)

Davidski said...

Alright, last set of trees for now.

I have run Goyet in the past. Very similar to Kostenki14 in TreeMix.

Alberto said...

On the trees with Indian_South and Israel_Natufian, the one with 3 edges looks absolutely standard and clean and simple. It's similar to the one with Karitiana.

I think I would go for that more simple explanation instead of other more complicated and speculative ones (given the two different models shown in the trees). But who knows, we really miss so much ancient DNA from Asia that everything is possible.

Davidski said...

The problem with Indians, including South Indians, is that they have very complex ancestry, so it's not always easy to tell what their position on a tree actually means, and why they're getting or producing migration edges.

So I'd only run them when I'm actually looking at South Asian population history. But in the above set of trees, I think two things are clear:

1) South Indians have at least one type of Iran Neolithic-related ancestry

2) Their ANE-related ancestry includes the same sort of stuff that the Han Chinese have

Rob said...


Thanks .
2 quick questions:
- do Papuans and Australoids have any of the southeast Asian ANE-like ?

- does the tree mix showing tha the borth Eurasian admixture in CHG coming from northwest Eurasian instead of ANE square with other tests ?

Can you specify which tree are you favouring and which you find more speculative ?

Alberto said...


From the ones here:

The 3rd one looks very standard, without any Basal/African into South/East Asians, or lots of ANE into all of South/East Asians too.

We just have a Basal Eurasian population first (Israel_Natufian) which gets 50% admixture from the base of the branch that leads to Villabruna/Bichon. Then we have East Eurasians on one side and West Eurasians on the other (with Ust-Ishim being in no-mans-land, somehow intermediate). Then West Eurasians split into the more western branch (Kostenki, Villabruna, Bichon) and the more eastern (Central Asian?) one (MA1-AG3 and Iran_N), with Iran_N getting 35% Basal Eurasian.

Finally, Indian_South gets 18% ANE.

So it's pretty much what people have been figuring out these last few years.

Davidski said...

Don't know about Papuans.

But Supplementary Information 11 in Lazaridis et al. 2016 is a useful read.

Rob said...

Thanks Dave & Alberto

Colin Welling said...


Wait, why are you calling this ANE, or even ANE like? Both of the edges in the first two tree's are so close to the divergence point of ANE and the Paleolithic European lineage which means these hypothetic migrants were barely any different from the the ancestors of paleolithic europeans.

I also have a very simple, technical, question. Does the migration edge also have a direction? Could it be that ANE has some ENA instead?

In regards to the meaning of it all. When did ANE actually evolve. Did it necessarily evolve around the time of MA1? If so is that a strong indication it evolved around siberia? What do you think the extent of ANE was around the time of MA1?

Do you think that East Asians have actual siberian, MA-like, heritage? Do you think Europeans have siberian, MA-like, heritage or could there MA like heritage be from a place other than siberia? The way you are talking about ANE in this post makes it should like ANE existed over a long period of time and could have extended over a large span of space.

I know that was a lot of questions but can you try to answer them

ryukendo kendow said...

David, thank you very much for the new trees, they are very illuminating.

@Alberto, just like to point out that except for tree no 3 in the set I requested, and Matt's single tree, all of the other trees, including all of the newest sets for me and Rob, uniformly paint a picture of reduced similarity of far southwestern and far eastern populations to populations in the centre of Eurasia. None of trees except 2 have ENA where they are supposed to be in the classical model, so we know that Treemix is trying to depict a phenomenon here.

Just thinking out loud, certain facts have already changed on the ground, for example now with the genomes from Fu et al, and the newest 6% ANE in East Asians in lazaridis, there are no longer *any* symmetries between members of the previous West Eurasian clades and East Eurasian clades, so except for the Kostenki-WHG clade, its no longer even clear whether any of the previous clades exist at all in their putative forms. In fact we get the situation of populations in a chain across Eurasia all sharing various levels of gene flow with each other and thus intermediate and asymmetrical to all other populations, e.g. ANE is now intermediate between the non-ENA admixed WHG (e.g. Bichon) and ENA:

Natufian/Iran_N<>WHG<>ANE<>Karitiana<>Han<>Onge<>Ust Ishim<>Papuan

There are various ways that Treemix can attempt to slice and dice this variation, for example, in the latest set they have all genomes from ANE to Papuan in a single clade as 'Crown' Eurasian', and ANE, Karitiana, Han, Onge having some interlacing to the exclusion of Papuan and U-I, but Natufian/Iran sending clades to WHG, i.e. Chad's scenario. Or they can have WHG, ANE and Karitiana+East Asians having some interlacing, and Basal in E Asians and another Basal in Natufian/Iran_N. Or they can have Natufian/Iran_N all the way to ANE as a single clade, then Basal comes into East Asians, which are reticulated with ANE.

All these asymmetries probably make the models very unconstrained with too many degrees of freedom for some consensus branching order to emerge. But what is clear is that populations at the centre of Eurasia have some extra connection each other to the exclusion of populations at the side, which treemix consistently demonstrates a preference to depict, whether caused by reticulation or clading or whatever.

ryukendo kendow said...

Hmm, just an example here, if we have A, B C, D populations, and A and B are equidistant and symmetrical to C and D, then its quite likely that the scenario (A, B) (C, D) is correct. If we have A, B, C, D and C and D are equidistant to A and B, but B is closer to C+D than A is, then the situation may be (A, (B, (C, D))). If we have A, B, C, D, and C and D are again equidistant to A but B and C are closer to each other than A and D are, then the tree is likely (A, B) (C, D) or (A (B (C, D))) but with B<---C. (This is the central insight which allows us to design f4 and D stats with populations in positions that give us useable information.) However if we have B intermediate between A and C, and C intermediate between B and D, then all comparisons are asymmetrical, no equidistances exist, and there is no sure solution that can be readily inferred, except some kind of generalised description of A<>B<>C<>D. The various intermediacies can be accounted for by either gene flow or differentiated branching, e.g. (A, B---)--(---->C, D), or (A, B<---)--(----C, D) or even (A, ( B, C--)-)----(--> D) or (D, ( C, B--)-)----(--> A), or even multiple pulses of admixture. This seems to be the situation for the populations we have in Eurasia now, except for the Kostenki-WHG clade, which still retains equidistances with outside groups, esp in the non-Villabruna populations. All other comparisons seem quite asymmetrical, or at least asymmetrical enough for Treemix not to produce robust clading and for multiple fits to be feasible. With the AG3 edges into later Villabruna, and GoyetQ similarities with Mal'ta, maybe even the symmetries of the Kostenki-WHG clade may be a coincidence and even that group is looking a bit wobbly.

Colin Welling said...


"Further, ANE seems to be a parallel to the non -basal components of U-I, Onge, ASI, the latter of which all have Basal, but ANE does not."

Yes, if you take consider proper ANE, i.e. the fully developed ANE that was found in MA and A.G.. What surprises me is the two branchings that would eventually lead to East Asians according to the model david is proposing. One branch split before the divergence of the Paleolithic Europe lineage (what someone called "basal" II) and on branch that split shortly after the divergence of the Paleolithic Europeans lineage (pre ANE).

We really need to come up with a name for populations at the base of a branch that is different from the population at the end of a branch. For example, we can call the base of ANE, relative to the paleolithic europeans, pre ANE. So pre ANE for short, and pre ANE (Paleolithic Europeans) to be technical.

human443 said...

All of these models run into one of two problems

1 - If both Han and Ust-Ishim are ~60% East Eurasian and ~40% ANE, then Ust-Ishim should be most related to Han, less related to ANE, less related to paleoeuros.

2 - If Ust-Ishim and Han share a node to the exclusion of ANE, as well as sharing roughly a third of some hyper basal ancestry, they should be much closer to each other than either is to ANE. This is not the case. No amount of ANE input into either would correct this.

'Ane layers' - Problem 1
'Ane and Ena as sister clades' - Problem 2
'Misc trees'
1 - problem 1
2 - problem 2
3 - Overall phylogeny looks excellent, covers basal and everything, only needs an arrow from ANE to Han/Onge to be complete.
4 - problem 1
5 - problem 2

German Dziebel said...


"Adding Karitiana makes the East Asians look unadmixed"

Aha! When you don't doctor Amerindians out, your newly found admixture between MA-1 and Han disappears and the software favors Amerindians vs. Han as more West Eurasian shifted. (As it did in the original study of MA-1 by Raghavan et al.) So, it's really the ancestors of modern Amerindians (=ANE) who are admixed into both West Eurasians and East Asians. This is an old pre-Ust-Ishim migration. Plus Han may be especially closer to Northern Amerindians because a later Holocene-age back migration from North America to East Asia (without Western Eurasia involved).

Chad Rohlfsen said...

Here's an interesting one. To make this work, I used my two Onge sets. One from the non-public Human Origins set, and another from a public set with Onge and Jarawa, with 8 million SNPs.


right pops:

best coefficients: 0.223 0.549 0.155 0.073

std. errors: 0.017 0.065 0.042 0.015

fixed pat wt dof chisq tail prob
0000 0 2 0.272 0.872706

Davidski said...

Colin & human4443,

I'm saying the admixture is ANE-related. Not actually ANE. The "related" concept is very flexible, as per the Laz et al. supp info...

Two notes of caution are necessary here. First, the admixing populations need not be necessarily “close” (either geographically or genetically) to the EHG and Onge, but they are in some sense related to them so that present-day Eastern Eurasian populations have intermediate allele frequencies between them. This is plausible given that both EHG and Onge show genetic affinities that stretch well beyond eastern Europe and the Andaman Islands, and may represent in some sense more widely dispersed populations. The EHG share more alleles with a ~24,000-year old Upper Paleolithic individual (MA1) from Siberia 6 than any other ancient or present-day population, and the Onge are representative of both “Ancestral South Indians” contributing to populations of the Indian subcontinent 7,8 but also of a minor stream of ancestry present in Amazonians 9 . Second, in contrast to western Eurasia there is currently no genome-wide ancient DNA data from eastern Eurasia. As an analogy, prior to the availability of genomes other than the Tyrolean Iceman, Europeans were modeled as a 2-way mix of “Sardinians/Iceman” and an “Ancient North Eurasian” ancestry from which the Native Americans were also descended. However, subsequent work with ancient DNA provided a much more detailed picture, involving a three-way mixture and more proximate source populations 1-3,10 . Thus, while our results demonstrate widespread ancient admixture in eastern Eurasia, the story of eastern Eurasian origins will doubtlessly be more complex than our model.

ryukendo kendow said...

Agree, the position of Ust-Ishim in this is the most difficult and conflicting element that most contradicts the stats we've seen so far.

@ David

David, do you mind running the following stats?

Chimp Ust_Ishim Han Bichon
Chimp Ust_Ishim Onge Bichon
Chimp Ust_Ishim Han Villabruna
Chimp Ust_Ishim Onge Villabruna
Chimp Iran_Neolithic Ust_Ishim Han
Chimp Israel_Natufian Ust_Ishim Han
Chimp Iran_Neolithic Ust_Ishim Onge
Chimp Israel_Natufian Ust_Ishim Onge

Ryan said...

What's the oldest sample we have from western Eurasia, and how do they fit in to this tree?

I think the edge from Africa into ENA give pretty good evidence for 2 migrations out of Africa, with the older migration persisting best in south/east Asia.

IIRC an edge from Africa to Melanesians in particular shows up pretty often, no?

Chad Rohlfsen said...


Did any TreeMix runs have a BE edge, plus the edge from MA1 to Ust Ishim?

Davidski said...

It seems that East Asians are close to 50/50, so it's either one or the other that is shown.

Iranocentrist said...

"other than the Tyrolean Iceman, Europeans were modeled as a 2-way mix of “Sardinians/Iceman” and an “Ancient North Eurasian"

Wasent it WHG instead of ANE ?

Davidski said...

It was Oetzi/Karitiana back in the day. You should read the paper cited from 2012.

human443 said...

@ Davidski,

Problem 1 could be minimized (although I doubt eliminated) by Ust-Ishim being composed of each of the two components on veeerryy short drift lengths from their initial splits, so short in fact, as to barely be justifiable in calling ENA or ANE.

Problem 2 is not addressed by your statement. Ust-Ishim could be composed of 20 different crown Eurasian branches spanning from Antarctica to Jupiter, but as long as it shares that substantial chunk of hyper basal admix with Han, it will be closer to Han than to anyone else.

Davidski said...


Chimp Ust_Ishim Han Bichon -0.0092 -1.762
Chimp Ust_Ishim Andamanese_Onge Bichon -0.0016 -0.288
Chimp Ust_Ishim Han Villabruna -0.0077 -1.457
Chimp Ust_Ishim Andamanese_Onge Villabruna 0.0004 0.069
Chimp Iran_Neolithic Ust_Ishim Han 0.0302 6.034
Chimp Israel_Natufian Ust_Ishim Han 0.0127 2.411
Chimp Iran_Neolithic Ust_Ishim Andamanese_Onge 0.017 3.217
Chimp Israel_Natufian Ust_Ishim Andamanese_Onge 0.0038 0.688


For one, TreeMix shows that Han have admixture that Ust_Ishim lacks. And two, there's only so much that these graphs can show clearly. So you shouldn't take the output too literally.

human443 said...


I do not contest admixture into Han to the exclusion to Ust_Ishim. I believe the most parsimonious model is that Ust_Ishim is an early East Eurasian, and that subsequent East Eurasians received varying degrees of input from an early branch of ANE.

And you are right, there is only so much these graphs can do, so their conclusions must be double checked...I did so, and the conclusions were illogical. Nothing overtly literal about it.

Chad Rohlfsen said...

I think the edge to the base of Ust Ishim is only to explain his position, but not to be taken literally. If Ust Ishim and Han were much more alike than they are to MA1, the stat Chimp Ust Ishim Han MA1 would be very significant and not so small. Checking Chimp Han Ust Ishim MA1/Onge should confirm this. Han may be closer to MA1 than to Ust Ishim, nullifying that edge. I think the Onge lack West Eurasian ancestry. They're quite divergent from other East Asians in relationship to MA1 and Papuans. Being only y D and mt M speaks to some extensive isolation.

human443 said...


Mbuti Han Ust_Ishim MA1 0.0183 3.828 40059 38616 683873
I don't have the other stat, but it would be even more significant, as all East Eurasians are closer to each other than to Ust-Ishim.

I don't think the Onge lack West Eurasian, but they definitely have less of it. Papuan look to have even less, but still a small proportion. (Papuan are what, 20% mainland East Asian however?). Australian aborigines seem to be the far end of this cline...
Mbuti Han Kostenki14 Ust_Ishim 0.0078 1.616
MbutiPygmy Papuan Kostenki14 Ust'-Ishim 0.0098 1.456
MbutiPygmy Australian Kostenki14 Ust'-Ishim 0.0144 2.061

Could somebody run a few more stats with Australian?

Mbuti Kostenki14 Australian Ust_Ishim
Mbuti MA1 Australian Ust_Ishim
Mbuti Australian Onge Han
Mbuti Ust_Ishim Australian Han
Mbuti Australian Ust_Ishim Han
Mbuti Australian Kostenki14 MA1

jparada said...

I think there's a problem with this scenario from uniparental markers, or at least a different one is somewhat more likely. That being, East Eurasians harbor some ancestry from a deeply divergin, pseudo-african lineage, which would be carriers of Y-dna haplogroup D, and a "Crown Eurasian" lineage, likely carrying Y-dnas K2 and C and mt-dna B. In this scenario, ANE would be a mix of this "Crown East Eurasian" lineage and something related to Palaeoeuropeans, thus explaining why ANE is closer to East Eurasians while carrying none of this "Basal East Eurasian" lineage.

jparada said...

(continued) and the presence of Y-dna K2 derived lineages in both ANE and East Eurasians, but not in Palaeoeuropeans.