Is there really Basal Eurasian, and or Iranian/Caucasus-related ancestry in Anatolia’s first hunters and farmers? Maybe not…

I have been having a go at some new graphs here to see if there is really any Basal Eurasian in early Anatolians. While I am limited to just a few samples and do not have access to Dzudzuana (Lazaridis et al (2018) yet, it has become quite clear that Basal Eurasian, or Ancient North African is not needed in these samples. And while it was also reported in Feldman et al (2019) that early farmers from Boncuklu harbored ancestry from a group related to Iranian farmers or Caucasus hunter-gatherers, this need not be the case.

While going back to look at d-stats between some of these groups, I noticed some oddities that were produced when using African populations and modern East Asians, with other ancient Eurasians. Modern Africans show a preference for more recent than ancient Eurasians. There’s even a significant preference for Chalcolithic Iranians in 10,000 year old samples from Malawi, which makes no sense. This is probably some type of artifact in the data. This, to me, was creating the impression of a Basal Eurasian branch where none need be found. When you compare ancient Eurasians to each other, it does disappear.

For instance, the Z-scores using Chimp as an outgroup, and even Altai or Denisovan, were completely different than using Mbuti, Mota, or any other African individual or population. The significance of these stats vanished when comparing the first Anatolians versus ancient European hunters.

Not only that, it also disappeared when using ancient East Asians and Native Americans such as Tianyuan, Vietnam_N, Vanuatu_2900BP and Brazil_LapaDoSanto_9600BP. Sure enough, when you compare modern East Asians versus these samples, the modern groups end up being significantly closer to Ust_Ishim and Africans. So, instead of all ancient East Asians having Basal Eurasian, or all modern East Asians having ancestry from Ust_Ishim or some special relationship with Africans, this may also be an artifact. The second is possible, but probably minimally so.

So, what we have is Dstats in the order of (Chimp/Altai/Denisovan, Ust_Ishim; Anatolia_HG/Boncuklu_N, SunghirIV/Kostenki14/Vestonice16/Tianyuan/Vietnam_N/Brazil_LapaDoSanto_9600BP) all being quite insignificant. This had me look into creating some more graphs to look at the issue and deciding to stay away from using modern East Asians and Africans in analysis of ancient Eurasians.

The graph that I ended up with combined Natufians, Iranian farmers from Ganj Dareh, both the Anatolian hunter-gatherer from Pinarbasi and early farmers from Boncuklu, Upper Paleolithic European SunghirIV, Iberomaurusians, and Upper Paleolithic samples from Siberia and China, Ust-Ishim, Yana, and Tianyuan, respectively.

For this graph, I grouped SunghirIV and the Anatolians on a branch together, separate from Ust_Ishim and Tianyuan, as neither show any signs of shared drifted with these groups that the other doesn’t have.



With this graph, it brings up the interesting prospect that deep ancestry did not reach Anatolia, or the Caucasus with Dzudzuana, but later in the Levant, Caucasus, and Zagros region. With the very minimal excess sharing between Boncuklu and Iran, I chose to use Boncuklu as the population that donates the Anatolian, or Dzudzuana-like stuff to Iran, while the Natufians preferred a source more like Pinarbasi (Anatolia_HG).

Without samples covering a range of time between 30,000-10,000 years ago across all of West Asia, it is hard to say if this will pan out. For now, this is the best that I can do.

While neither the Anatolians or Sunghir share any special relationship with Tianyuan or Ust-Ishim, Yana is significantly closer to Upper Paleolithic Europeans, suggesting they stem from the same branch.

Iberomaurusians fit best as a mix of the so-called Basal Eurasian or Ancient North African and something coming from the same branch as, but before the split of Boncuklu and Anatolia_HG.

The Iranian farmers first wanted to be grouped with Boncuklu, then onto Yana and Iberomaurusians. Afterwards, there was a Z-score between 2-3 that asked for more ancestry related to Tianyuan, but deep than the branch to Yana, going into Iranians. I had noticed this long ago and included it in previous posts. It does keep the graph together much better as more populations are added as well.

As previously mentioned, I avoided the use of any African or modern East Asian population. I did not want to have any artifacts create unnecessary admixture events or ghost populations.

For the next graph, I added in the Gravettian samples from Vestonice. The graph still held together and with an insignificant Z-score.


For my final go, I decided to add the Epigravettian sample, Villabruna, a 14,000 year old hunter-gatherer from Italy and the reason behind the name “Villabruna cluster”. Villabruna actually shares significantly more drift with ancient Anatolians than Vestonice does, but also significantly more with Ancient North Eurasians (ANE) like Yana and MA1, than the Anatolians do.

For this reason, I chose to make Villabruna, the much younger sample, a mix of Gravettian, Anatolian, and Yana. Due to the fact that there is really no place for some pure population to hide between Dzudzuana and SE European Gravettians, along with the fact that R1b and significant relationships with Native Americans are there, it makes sense that ANE would also be in there. To me, this model feels much more parsimonious than trying to make everyone some mixture involving the much younger population that has no evidence of existing before a much later period. As I said, it is running out of real estate for some place to hide-out for 20-30,000 years.

The major part of the reason I think that this Basal Eurasian or North African was needed in Anatolians, was making them have significant ancestry from Villabruna, who has that relationship with ANE and East Asians over Anatolia. This means that you then must compensate for that relationship not in Anatolia by giving them deeper ancestry that is probably not there. They share enough drift that these tools allow you to branch them together, but then you get caught up in a catch 22 by having to add all of these extra admixture events. For instance, having to make Anatolia a mix of Natufians/Levant, Iran, and WHG, rather than just another branch of West Eurasians that contributes to North Africans, Mesolithic Europeans, and into the Iranian plateau and maybe as far as India before the Neolithic.

In this last graph, Villabruna is a mix of  77% Vestonice and Anatolia_HG (51% and 49% respectively) and 23% from a lineage related to Yana.


While there may be new data that makes this impossible, there is certainly nothing there right now to say that it isn’t. There isn’t a single stat out there with ancient Eurasians that can break this tree down as far as I know. I included many populations to make sure that it would hold up. I look forward to seeing more data out of West Asia, and particularly access to the Dzudzuana samples to see if this holds up.


Just to make things a little more clear, I will show in more detail what I am talking about. It is the use of Africans and modern East Asians that gets in the way of robust analysis of ancient Eurasians. For whatever reason, there is an artifact that draws modern Asians and ancient West Asians much closer to Africans than UP samples from Europe and Asia.

Here are a few d-stats to show what I am talking about..

Chimp Malawi_Ho IBM Ganj_Dareh -0.00011 -0.321 580681
Chimp Malawi_Holocene IBM SunghirIV -0.001876 -4.17 584427
Chimp Malawi_Holocene IBM AnatoliaHG -0.000176 -0.411 507596
Chimp Malawi_Holocene IBM Iran_ChL 0.000235 0.654 569410
Chimp Malawi_Holocene IBM Yana -0.002173 -5.98 589155
Chimp Malawi_Holocene Ganj_Dareh SunghirIV -0.001849 -4.311 581183
Chimp Malawi_Holocene Ganj_Dareh AnatoliaHG -0.000137 -0.33 506101
Chimp Malawi_Holocene Ganj_Dareh Iran_ChL 0.00042 1.45 567512
Chimp Malawi_Holocene Ganj_Dareh Yana -0.002128 -6.497 585873
Chimp Malawi_Holocene SunghirIV AnatoliaHG 0.001456 2.912 507356
Chimp Malawi_Holocene SunghirIV Iran_ChL 0.002128 5.01 569380
Chimp Malawi_Holocene SunghirIV Yana -0.000277 -0.64 591351
Chimp Malawi_Holocene Anatolia_HG Iran_ChL 0.000391 0.959 497265
Chimp Malawi_Holocene Anatolia_HG Yana -0.001986 -4.528 511289
Chimp Malawi_Holocene Iran_ChL Yana -0.002501 -7.398 573952
Chimp MbutiSGDP IBM SunghirIV -0.001592 -4.869 1037166
Chimp MbutiSGDP IBM AnatoliaHG -0.000818 -2.706 836965
Chimp MbutiSGDP Ganj_Dareh SunghirIV -0.001268 -4.134 1003826
Chimp MbutiSGDP Ganj_Dareh AnatoliaHG -0.000363 -1.268 823271
Chimp Mbuti.SGDP SunghirIV AnatoliaHG 0.000712 2.001 847590
Chimp Ust_Ishim SunghirIV AnatoliaHG 0.000044 0.062 845798
MbutiSGDP Ust_Ishim SunghirIV AnatoliaHG -0.000853 -1.281 882870
Tianyuan Ust_Ishim SunghirIV AnatoliaHG 0.000254 0.312 705443

As you can see, there are some things here that seem a little off. It is even worse when you use samples like Ust_Ishim, Yana, and Tianyuan. I can say that it isn’t about there being some type of ancient attraction to Chimp either. The attraction is with Africans. This isn’t only the case for Mbuti, but even ancient Africans like Malawi_Hora_Holocene. Stats are even more significant for the Chalcolithic Iranians, which makes zero sense for 10,000 year old samples from SE Africa. The trees below shows that when you combine Mbuti to here, it doesn’t mess up the Eurasian part of the tree, but requires admixture from those groups into the Mbuti. The first tree is just adding Mbuti to the base of the tree, then the admixture to the Mbuti in the second graph.



The graph above would be even worse if I added a modern East Asian like the Onge. Ths would require Onge admixture to the Mbuti too. All-in-all, I think this shows that if you include an African or modern Asian in your graph to start, you are already setting yourself up for issues here that might affect the outcome. While Chimp is maybe not completely ideal as an outgroup, it is more inline with the graph that has no African or Chimp in it. The stats line up much better with the graph. Not only that, it is apparent that any deep ancestry not covered within in the Ust_Ishim, West, East branching is still easily detectable without Chimp or an African outgroup in the graph.

Another little interesting thing with this graph that lines up with the stats is the following comparison between Natufians and Iranians in relation to Yana and Tianyuan. If all East Asian ancestry in Iran was mediated via a Yana-like population, then the stat with Tianyuan should be minimal compared to Yana. However, this isn’t the case and lines up nice with the graph asking for significant East Asian into Iran.

Chimp Tianyuan Natufian Ganj_Dareh 0.00292 4.92 414476
Chimp Yana Natufian Ganj_Dareh 0.00139 2.683 484580

I think this all shows it is definitely safe to run qpGraph with no African or Chimp as an outgroup and running qpGraph with outpop: NULL. Safer, it seems, than including said groups at the risk of affecting analyses.


Felman et al (2019) Late Pleistocene human genome suggests a local origin for the first farmrs of central Anatolia.

Lazaridis et al (2018) Paleolithic DNA from the Caucasus reveals core of West Eurasian ancestry.


The big picture for West Asia, and specifically Anatolia, after Pinarbasi

Thanks to Feldman et al (2019), we got our first look at the ancient Anatolians from the Pleistocene. Due to the strong cultural links with later peoples, such as Boncuklu, I kind of figured that the hunter-gatherers would be pretty similar to the first farmers. The piece by Baird et al (2013), “Juniper Smoke, Skulls, and Wolves’ Tails“, was very insightful and does hint at continuity in the region.



To take a look at the formation of the first farmers of West Asia, I used the samples from Feldman et al (2019), Lazaridis et al (2016), and van de Loosdrecht et al (2018) to see what kind of picture emerges.

This first graph, is a simple tree that involves the formation of Boncuklu_N and Anatolia_N (Barcin), while using farmers from Iran and the Levant as additional sources, besides Anatolia_HG (Pinarbasi).


qpGraph Models for the Anatolian Farmers

The first graph is a simple set up of the base populations that are the most likely contributors to the formation of Boncuklu and Barcin. While there may be a better source in Eastern Anatolia or Syria, these will do for now.


The addition of Boncuklu to the graph led to a worst-Z score approaching five, asking for admixture between Ganj_Dareh_N and Boncuklu_N.


The inclusion of admixture from Ganj_Dareh_N to Boncuklu brought the Z-score down below three and formed a good base for Barcin, and is in general agreement with the findings of Feldman et al (2019).


While Anatolia_N is significantly closer to Boncuklu than any of the other populations, there is still a significant relationship with Levant_N shared only with the Barcin farmers.


This next graph shows more agreement with Feldman et al (2019), with significant Levant_N ancestry being found in the Barcin farmers. However, there is still a significant relationship between Ganj_Dareh and Barcin that is not resolved.


This last graph did bring the worst Z-score to a more acceptable level, around three. This does differ a bit from qpAdm, where you can create a successful model without any Iranian admixture on top of what is found in Boncuklu. The Iranian admixture is essentially at the level of the standard error, making it unnecessary.



qpGraph Models for the Mixing Between Populations in North Africa and West Asia from 15,000-10,000 Years Ago

To begin this simple graph, I started with the older samples from each region, with Natufians being a mix between Iberomaurusians and ancient Anatolians, which does fit fine and first demonstrated by Lazaridis et al (2018)(pre-print). Since I do not have the Dzudzuana genomes, I used Anatolia_HG (Pinarbasi) as the stand-in for that admixture into the Iberomaurusians (IBM).


For the next graph, I started by adding Levant_N to a branch related to Natufians. I didn’t think starting with Levant_N before Boncuklu_N is a big deal as several of the samples are contemporary and one need not come from the other. The most significant score left from this graph asks for a stronger relationship between Ganj_Dareh and the Levantine farmers.


After adding an admixture event between Natufians and Ganj_Dareh to create Levant_N, there is still a need for a stronger relationship between Anatolia_HG and Levant_N.


This next graph only required minimal admixture from Anatolia_HG (five percent), to bring down the Z-score below 2.


For the last tree, I added in Boncuklu and also needed the admixture from Ganj_Dareh to Boncuklu to bring the Z-score down. In this last graph, Anatolia_HG ancestry does overtake Ganj_Dareh ancestry in the Levant_N samples. Anatolia_HG-related ancestry also is the most widespread component in Anatolia, the Levant, and North Africa.



Full qpGraph of Ancient Farmers 15,000-10000 Years Ago

For the last set of graphs, I included GoyetQ116-1 and MA1 to cover ancient European to Siberian ancestry.  Ust_Ishim was added to try and figure out the amount of Basal Eurasian that would be needed for each sample. I also included the Onge for any additional East Eurasian ancestry that might be needed.

As a base, I started with just Anatolia_HG, IBM (Iberomaurusians), and Natufians.


For the next graph, I included Ganj_Dareh_N as a mix of MA1 and Basal Eurasian. This left a very significant Z-score between the Onge and Ganj_Dareh_N, suggesting additional East Eurasian ancestry is needed for Ganj_Dareh_N.


For this graph, an additional 29 percent admixture from Onge to Ganj_Dareh_N helped bring down the worst Z-score significantly.


This next graph includes Boncuklu_N, coming from a branch related to Anatolia_HG. The worst Z-score is also suggestive of gene-flow from a population related to Ganj_Dareh_N into the Boncuklu_N farmers.


With admixture from Ganj_Dareh_N to Boncuklu, the worst Z-score has dropped to near 3. The next addition to the graph will be Levant_N, from a branch related to Natufians.


With the inclusion of Levant_N, admixture is again needed from Ganj_Dareh, to Levant_N, just as with the simpler graphs.


With the addition of Ganj_Dareh admixture, the Levantine farmers next needed some extra ancestry from a group related to Anatolia_HG.


This last graph shows what is an effective model for the deeper ancestry and mixture between ancient North African and West Asian populations.




From what I have been able to put together, there is an agreement with the findings of Feldman et al (2019) when it comes to the formation of Anatolian farmers and the importance of the Pleistocene hunters, or a related group, to the formation of not only Anatolian farmers but also Levant_N. The only difference with qpGraph is that it does want additional ancestry from Ganj_Dareh_N into the Barcin farmers. To re-iterate Feldman et al (2019), there is the appearance of great continuity through the development of agriculture in Anatolia, but this could change with more samples, particularly Eastern Anatolia and Syria.



Baird et al. (2013) Juniper smoke, skull sand wolves’ tails. The Epipaleolithic of the Anatolian plateau in its South-west Asian context; insights from Pinarbasi.

Feldman et al. (2019) Late Pleistocene human genome suggests a local origin for the first farmers of central Anatolia.

Lazaridis et al. (2016) Genomic insights into the origin of farming in the ancient Near East.

Lazaridis et al. (2018) (pre-print) Paleolithic DNA from the Caucasus reveals the core of West Eurasian ancestry.

van de Loosdrecht et al (2018) Pleistocene North African genomes link Near Eastern and sub-Saharan African human populations.




A Working Tree for Ancient Eurasia

For quite some time I’ve been a proponent of early entry of Basal Eurasian into European hunter-gatherers. This looked a little clearer with the likes of the Iron Gates hunters from Mathieson et al (2018). Not only that, but I was noticing some slight shift from the Gravettians from GoyetQ116-1 (Fu et al, 2016), along with the West Eurasian side of ancient Native Americans, such as the Clovis boy (Rasmussen et al, 2014). While this became much more possible with the discovery and modeling of European hunters with the Dzudzuana pre-print (Lazaridis et al, 2018), I also noticed that another, more ancient Eurasian looked to have a good amount of this deep lineage commonly referred to as Basal Eurasian. This sample was Sunghir IV from Sikora et al (2017).

The Sunghir site is one of the most famous and well-studied sites in Upper Paleolithic Europe. The child burials have been subject of much discussion and even VR reconstruction efforts for their faces have produced interesting results (Holger, 2017).



The interesting thing about SunghirIV is that it was a partial femur that was packed with ochre (Sikora et al, 2017). Not only that, the bone chemistry of this individual pointed to a different geographic place of origin for this individual compared to the other Sunghir remains (Sikora et al, 2017).

Looking at D-stats to Form the Base of the Tree

Several things jumped out at me looking at ancient East Eurasians, both Tianyuan (Yang et al, 2017), and Vanuatu_2900BP_all (Lipson et al, 2018), had the same relationship to ancient West Eurasians and to Ust-Ishim (Fu et al, 2014). Not only that, GoyetQ116-1 and MA1 both shared the same relationship with Ust-Ishim and ancient East Eurasians, unlike the other West Eurasian samples. This suggests that at one time, the from at least 36,000 years ago, there may have existed a huge meta-population stretching from Western Europe, all the way to at least Lake Baikal.

Out Target X Y D-stat Z-Score SNPs
Chimp Tianyuan GoyetQ116 MA1 0.000116 0.132 455312
Chimp Vanuatu GoyetQ116 MA1 0.000022 0.026 276242
Out Target X Y D-stat Z-Score SNPs
Chimp Ust_Ishim GoyetQ116 MA1 -0.000697 -0.864 528039

Not only that, but Tianyuan must’ve branched off very early from the population that leads to both Vanuatu_2900BP_all and the ENA branch mixing into Native Americans, as Vanuatu is significantly closer to Clovis, but Tianyuan is not significantly closer to Vanuatu than Clovis, despite the latter having significant West Eurasian ancestry.

Out Target X Y D-stat Z-Score SNPs
Chimp Tianyuan Vanuatu Clovis -0.001261 -1.9 386513

Still, in that stretch of nearly 37,000 years, it appears there was no flow from West Eurasians into South East Asians.

The next thing I examined was Ust-Ishim’s place within this phylogeny, as it has been said that he was equally related to modern East Eurasians and ancient West Eurasians. However, a different story emerges when looking at ancient East and West Eurasians together. Ust-Ishim becomes a “Basal West Eurasian”, related to the branch that then mixes with ENA to form the clade shared by Goyet and MA1. While not as significantly set in the West Eurasians as Tianyuan is to East Eurasians, it is still the best fit and may be simply that Tianyuan may have that extra 5,000 years of drift from the population split in Eurasia.

Out Target X Y D-stat Z-Score SNPs
Chimp Ust_Ishim GoyetQ116 Tianyuan -0.001916 -2.264 624072
Chimp Ust_Ishim GoyetQ116 Vanuatu -0.001875 -2.23 373102
Chimp Ust_Ishim Tianyuan Vanuatu 0.000375 0.521 385677

Ust-Ishim being closer to West Eurasians, yet West Eurasians (Goyet and MA1) being much closer to East Eurasians than Ust-Ishim led me believe that admixture from an ENA group to a branch descended from a sister clade to Ust-Ishim was the best fit for MA1 and Goyet.

Out Target X Y D-stat Z-Score SNPs
Chimp Tianyuan Ust_Ishim GoyetQ116 0.002363 2.787 624072
Chimp Tianyuan Ust_Ishim MA1 0.002601 3.071 583856
Chimp Vanuatu Ust_Ishim GoyetQ116 0.002612 3.261 373102
Chimp Vanuatu Ust_Ishim MA1 0.002764 3.437 318634

The final piece to the puzzle was to figure out just how SunghirIV fit into this phylogeny. As someone that was not only significantly further from East Eurasians, but also Ust-Ishim, the only thing that seemed to fit this nicely would be Basal Eurasian ancestry.

Out Target X Y D-stat Z-Score SNPs
Chimp Ust_Ishim GoyetQ116 SunghirIV -0.002626 -3.37 724721
Chimp Tianyuan GoyetQ116 SunghirIV -0.00331 -4.132 620480

The only problem that I knew I would face is the confounding factor of East Eurasians being further from Ust-Ishim than Goyet and most of the ancestry in SunghirIV. Although I can see it as a confirmation of sorts of ancestry deeper than ENA into Sunghir, the worst Z-scores are not going to really reflect this but will probably make SunghirIV an island to himself, or involve ENA pops being closer and another West Eurasian being closer to Ust-Ishim.

Out Target X Y D-stat Z-Score SNPs
Chimp Ust_Ishim Tianyuan SunghirIV -0.000498 -0.608 803562
Chimp Ust_Ishim Vanuatu SunghirIV -0.000888 -1.173 430030

D-Stats Showing Relationships Between Ancient “West Eurasians”

The D-stats below show the more outlier status of SunghirIV and the more intermediary position of other ancient Europeans and MA1.

Out Target X Y D-stat Z-score SNPs
Chimp Ust_Ishim GoyetQ116 SunghirIV -0.002626 -3.37 724721
Chimp Ust_Ishim GoyetQ116 SunghirIII -0.001442 -1.983 670463
Chimp Ust_Ishim GoyetQ116 SunghirII -0.001702 -2.339 726072
Chimp Ust_Ishim GoyetQ116 SunghirI -0.001549 -1.946 534517
Chimp Ust_Ishim GoyetQ116 Vestonice16 -0.001394 -1.803 561487
Chimp Ust_Ishim GoyetQ116 Kostenki14 -0.001386 -1.794 702161
Chimp Ust_Ishim GoyetQ116 MA1 -0.000697 -0.864 528039
Chimp Ust_Ishim SunghirIV SunghirIII 0.001054 1.655 998961
Chimp Ust_Ishim SunghirIV SunghirII 0.00094 1.444 1080491
Chimp Ust_Ishim SunghirIV SunghirI 0.001261 1.897 790860
Chimp Ust_Ishim SunghirIV Vestonice16 0.001254 1.714 694500
Chimp Ust_Ishim SunghirIV Kostenki14 0.001499 2.178 1012073
Chimp Ust_Ishim SunghirIV MA1 0.002065 2.652 763784
Chimp Tianyuan GoyetQ116 SunghirIV -0.00331 -4.132 620480
Chimp Tianyuan GoyetQ116 SunghirIII -0.001474 -1.873 574387
Chimp Tianyuan GoyetQ116 SunghirII -0.001893 -2.351 621729
Chimp Tianyuan GoyetQ116 SunghirI -0.001434 -1.738 458764
Chimp Tianyuan GoyetQ116 Vestonice16 -0.002817 -3.472 508391
Chimp Tianyuan GoyetQ116 Kostenki14 -0.003163 -3.926 601682
Chimp Tianyuan GoyetQ116 MA1 0.000116 0.132 455312
Chimp Tianyuan SunghirIV SunghirIII 0.002014 3.032 739525
Chimp Tianyuan SunghirIV SunghirII 0.001629 2.332 800109
Chimp Tianyuan SunghirIV SunghirI 0.001864 2.606 590273
Chimp Tianyuan SunghirIV Vestonice16 0.000527 0.713 607772
Chimp Tianyuan SunghirIV Kostenki14 0.000656 0.921 777500
Chimp Tianyuan SunghirIV MA1 0.003533 4.299 580265

Tree Building in qpGraph

The first run here grouped MA1 and GoyetQ116-1 together, since both showed a similar relationship to East Asians and Ust-Ishim. Clovis was grouped with Vanuatu_2900BP_all, since there was significantly more drift shared with this pop, versus Tianyuan. Ust-Ishim was also placed at a position basal to West Eurasians due to the minor shift towards West Eurasians, versus East Eurasians. SunghirIV and the combination of SunghirI, SunghirII, and SunghirIII were placed together due to their closer relationship as well.


Since the worst Z-score from this run involved flow between Clovis and MA1, the next step was to create admixture from a branch related to MA1 into Clovis.


As this graph left a worst Z-score require MA1 to be closer to Vanuatu than SunghirIV, I decided to run an admixture edge from a branch related to Vanuatu into West Eurasians, post-Ust-Ishim, due to Goyet and MA1 being equally related to East Eurasians.


After this admixture edge, there was still a need for MA1 and likely Goyet using D-stats, so I decided to make a “Basal” branch leading into the Sunghir samples.


This next run asked for either more admixture from East Asians to MA1 and maybe Goyet, or a closer relationship between Goyet and Sunghir I-III. So, I decided that an extra relationship between Sunghir I-III was more appropriate here since there is no significant relationship between MA1 and Vanuatu compared to the others.


Interestingly, this run left a worst Z-score that asked for a closer relationship to Goyet, for SunghirIV. I interpreted this, along with the 0 drift edge on the branch to Goyet and MA1, that Goyet needed to be moved to a more basal position, relative to SunghirIV.


With this last tree, there is a good drop in the worst Z-score, to where the worst Z-score is now below 3. The relationship left may ask for an even more ancient admixture from an Ust-Ishim-related population into the ancestors of Clovis, but for this exercise, I didn’t find it necessary to capture what I was looking for.

Another qpGraph looking at SunghirIV

I took a couple more looks at SunghirIV, compared to Goyet and the other Sunghir samples, this time including Iberomaurusians, and with another one, the Pinarbasi hunter-gatherer. This first tree includes just the Iberomaurusians, where SunghirIV does ask for deeper ancestry, this shared with Iberomaurusians.


This last graph shows deep “basal ancestry” shared between IBM, Anatolia, and SunghirIV, with SunghirIV being intermediate between Anatolia and Goyet, and SunghirI-III being intermediate with SunghirIV and Goyet. In agreement with D-stats and other trees.



While the position of Ust-Ishim and whether or not Asians branch before them does need to be resolved, the position of SunghirIV and the fact that every other West Eurasian is intermediary points to a possible entry of Basal Eurasian to Europe before the time of Dzudzuana. More Aurignacian samples will help to clarify if Goyet is just an outlier, contains some artifact, or is representative of that population. The position of East Asians, relative to Ust-Ishim could be caused by some artifact in the data of ancient East Asians or modern East Asians.

It is also known that mixing modern and ancient samples can also cause issues as there is not only ancient attraction, but potential issues with the mapping, and haploid versus diploid. As ancient samples from East Asia, Europe, and hopefully West Asia come in, this tree may need some tweaking or see some confirmation.



Fu et al (2014) Genome sequence of a 45,000-year-old modern human from Western Siberia.

Fu et al (2016) The genetic history of Ice Age Europe.

Holger (2017) This VR app reconstructs 30,000-year-old homo sapiens faces in 3D.

Lipson et al (2018) Population turnover in remove Oceania shortly after initial settlement.

Rasmussen et al (2014) The genome of a Late Pleistocene human from a Clovis burial site in Western Montana.

Sikora et al (2017) Ancient genomes show social and reproductive behavior of Upper Paleolithic foragers.

Yang et al (2017) 40,000-year-old individual from Asia provides insight into early population structure in Eurasia.





The spread of the Afro-Asiatic language family is an argument that has several viewpoints. Some might say that it originated within Northeast Africa, and some that think it spread from the Near East.

Skoglund et al (2017) brought us our first East African pastoralist from Tanzania. This group is thought to come from pastoralists from Southern Kenya (Skoglund et al, 2017). In the paper, she was modeled as having ancestry from Levantine PPN farmers, as well as Mota, and the Dinka in another run (Skoglund et al, 2017).

With all this in mind, I decided to try a few different runs myself, including several ancient pops. This included the Levant_N samples, Levant_BA, Iberia_EN (As V88 in North Africa may have a Neolithic origin), Malawi_Chencherere_5200BP, Mota, Iberomaurusians, and South_Africa_2000BP.  I also used the new v37_1240K_HumanOrigins set from Harvard for another run.

qpGraph models of Tanzania_Luxmanda_3100BP

This first graph here is a little more complex, with several populations being used. It would appear that most of the ancestry of Tanzania_Luxmanda_3100BP comes from the Levant_BA, and a source related to the Malawi_Chencherere_5200BP samples which may represent mostly local ancestry from Tanzania.


This next set-up included most of the samples and had the initial most-significant, worst Z-score with Iberia_EN. Branching them off of Iberia made me have to look for negative statistics with Africans and it built from there, eventually requiring good admixture from the BA samples from Jordan. Even with the Iberian-related admixture, Jordanian-related input was still the most-significant single source. Interestingly, South African input is still required in a decent amount.


qpGraph Models Using the New V37 Set

The last two graphs here were done using the new v37 set from Harvard. I also included South_Africa_2100BP.SG samples in this first run. This more simplified graph from the v37 set was even more significant in the amount of input from a source related to the EBA samples from Jordan, and South African-related input was still strong.


For this last graph on the v37 set, I switched back to South_Africa_2000BP to see if there were any differences. In this run, the amount of Levant_BA-related input returned to levels more consistent to the other runs, but South African-related input dipped significantly in exchange for more from a source related to Mota.


qpGraph modeling of Egypt_New_Kingdom

Just out of curiosity, I decided to look at the oldest Egyptian samples from the New Kingdom period. This sample is of very low coverage, so extreme caution is urged on the interpretation of this output. Once again, there was a need for European farmer-related input to account for the ancestry in this sample. Also of interest was the fact that Dinka was a preferred population for Sub-Saharan input, over Mota. This may have interesting implications as a minimum date for the appears of N-S speakers into the Nile valley, but again, caution is needed here. In this run, Levant_N does make an appearance, along with a very significant input from a source like the BA samples from Jordan.



While I am getting differences in the West Eurasian input than was found in Skoglund et al (2017), the sources are still very related and I am also using the full 1240K and V37 sets, versus using a more reduced set from panel four and five from the Human Origins set. I think that input from a source that could have both Levant_N and Levant_BA input is quite plausible, and until we have Neolithic genomes from Egypt, it is hard to say how real any request for ancestry from European farmers is. The entrance of V88 to Africa is still a mystery, for now.

The large preference for input from a source more like Jordan EBA, rather than Levant_N may have interesting implications in the spread of Afro-Asiatic, but, as always, more sampling is needed before jumping on the bandwagon here.

As I work forward and dig deeper into Africa, I will possibly revisit this post and see how any output changes. I may also search for outgroups that were not used in Skoglund et al (2017), to see how any output I get from qpAdm may differ from the paper. With any luck, more samples from ancient Africa will help give answers to some of these questions, rather than raise many more.



Skoglund et al (2017) Reconstructing Prehistoric African Population Structure.


Potential extra Iberomaurusian-related gene flow into European farmers

Thanks to van de Loosdrecht et al (2018) we were able to see the ancient Iberomaurusian, or Oranian people. Not only that, but we see genetic continuity into the Capsian culture, from Fregel et al (2018). I think this is very important to consider as an option for a potential source of this gene flow from North Africa.

I believe this is important due to the fact that Iberomaurusians contain Y-DNA E-M78, and a precursor to E-V13, which is often seen by people as originating in Europe. How it got there has been the big question, and I think this may be the answer. Not only that, there could also be some female-based flow considering that there is J1c, T2b, and H1 among Iberomaurusians (Kefi et al, 2016) and European farmers, but not in Anatolian farmers.

As far as ancient samples, there is E-M78 among Sopot and Lengyel remains, as well as E-13 in Croatia Cardial and the Iberian Cardial. With that in mind, it seems that a mixing in the Balkans and disbursement from there is a more probable option.

I first performed D-stats with European and Anatolian farmers to see if there were any stats that jumped out. While these stats are not largely significant, it does seem that additional flow from Levant_N is not the source of this extra affinity. In fact, even within groups of these farmers, some had Z-scores > 3, in extra affinity to Iberomaurusians, without any significant score with Levant_N. The Bulgarian Early Neolithic samples I0704, I0706, I1298, and I2521 from Mathieson et al (2018) appear as the group with the least affinity to Iberomaurusians, without losing any affinity to Levant_N as well. There are some farmers that will approach or go over a Z-score of 3, but they are the minority. (The relevant D-stat sheet can be found here: )

qpGraph models for input to Cardial Impressa

The first sample that I looked at was Croatia_Impressa 1 and 2, which are samples I5071 and I5072, respectively. The first look here is with Croatia_Impressa I5071. This output showed that this sample can be modeled effectively as deriving about 97 percent of her ancestry from a Bulgaria_N-like source, and another 3 percent from Iberomaurusians.


This next sample, I5072 was also effectively modeled in the same way as I5071, with 3 percent input from a source like the Iberomaurusians.


qpGraph models for input to Iberia_EN

For Iberia, things are more complicated, with WHG input to these samples. Using Levant_N seemed to help in samples with extra WHG, so that they do not end up becoming something like 40 percent Bulgaria_N, 25 percent Iberomaurusian, and 35 percent WHG, or Iron_Gates. The two samples that I used here, Iberia_EN1 and Iberia_EN2, are samples I0409 and I0410.

The first graph here is for Iberia_EN I0409:


The next graph is for sample Iberia_EN I0410:


qpGraph models for input to LBK_Austria

The samples chosen here, LBK_Austria2, LBK_Austria4, and LBK_Austria7 are samples I5069, I5204 and I5207. Since they have some extra affinity to Iron_Gates, they were also modeled in the same way as Iberia_EN. I5069 is the first sample that I graphed.


Next, is the graph for I5204:


The last graph is for I5207:


These three samples, with significant and near significant scores towards Iberomaurusians, are all three pretty similar in ancestry proportions as well.

qpGraph models for input to LBK

The LBK samples that stood out are LBK2, LBK4, and LBK6. These are samples I0022, I0026, and I0048. The first graph is for sample I0022.


Next, is I0026:


Lastly, is LBK I0048:


qpGraph model for input to TDLN (Lengyel and Sopot samples)

The sample that stuck out for analysis was TDLN3, or sample I1891.


While not having any significant input, the appearance of E-M78 is for some flow from a source related to or more enriched for Iberomaurusian ancestry.

qpGraph models for input to Sopot_LN

The sample used here is Sopot_LN2, or sample I4168.




While these results are providing nothing that one might consider mind-blowing, it does open up a few options. With the only ancient instances of E-M78 being in ancient NW Africa, it is hard to say what was going on between there and the Levant in the Neolithic, as pastoral societies began to emerge. Direct flow from around Morocco or Algeria to the Balkans might be a bit harder to imagine that coming from potentially earlier Neolithic groups in Libya or possibly Egypt.

There is also the chance that the Levantines that mixed into the ancestors of European farmers may have had extra contacts with North Africans before going to Anatolia. Not having an ancient population that looks like a good fit as a direct source to Europe makes this a little harder to decipher.

The clear lack of Y-DNA E throughout Anatolia, but not in Europe is suggestive of extra contacts. The D-stats and graphs to point to potential contacts between the early farmers of Europe and North Africans. Nothing is really conclusive here with regards to how that flow happened, but the uniparentals and specific drift with Iberomaurusians came from somewhere outside of current potential sources, other than the Iberomaurusians themselves.



Fregel et al (2018) Ancient genomes from North Africa evidence prehistoric migrations to the Maghreb from both the Levant and Europe.

Kefi et al (2016) On the origin of Iberomaurusians: new data based on ancient mitochondrial DNA and phylogenetic analysis of Afalou and Taforalt populations.

Lazaridis et al (2016) Genomic insights into the origin of farming in the ancient Near East.

Mathieson et al (2018) The genomic history of Southeastern Europe.

van de Loosdrecht et al (2018) Pleistocene North African genomes link Near Eastern and sub-Saharan African human populations.

Deep tree with a branch related to Mota as “Basal Eurasian”

This is a short post here, before moving onto a larger one with something I find interesting regarding potential flow from Iberomaurusian to European farmers, maybe via the Capsian culture.  I will save anymore on this for that specific post.

I wanted to try a little something that we see often in papers of the couple years, where Mota is used as a stand-in for an ancestral branch called “Basal Eurasian” in modeling on qpAdm.  I don’t recall seeing this in any papers, so I thought that I would give it a try. These are three simple trees, the first involves Chimp, Mota, Ust_Ishim, IBM (Iberomaurusians), Anatolia_HG (Pinarbasi), Kostenki14, Tianyuan, and Natufians. Theses samples come from Lazaridis et al (2016), van de Loosdrecht et al (2018), and Feldman et al (2019).

Technically, I didn’t intend to use a Basal branch, and thought that if anything was needed between Mota and Ust_Ishim, the worst Z-score would require something away from Mota’s branch and diverging instead on the one towards Ust_Ishim, which I should expect if my interpretation of the tree is correct. While I do not want to speculate on Mota’s specific place in tree between “African” and “Eurasian”, nor his potential non-“African” ancestry (I have serious doubts about being able to decipher that after looking at IBM and also potential artifacts with regard to stats involving African samples), I wanted to focus more on seeing if direct ancestry from a related source does work. The results were interesting and did work, just as they do in qpAdm.

While running this, I was curious to see if I could avoid the Z-score between Natufians and IBM with a direct contribution from Mota. This still left a very significant Z-score, showing that direct contribution from a source related to Iberomaurusians was going to be the only way forward.

This was pretty close to the input from both IBM and Dzudzuana, from Lazaridis et al (2018). I eagerly await that sample, to see this a little more. I also wondered about the low Mota to Anatolia edge; thinking it has something to do with what I believe is Kostenki14 already possessing some Dzudzuana-related ancestry, compared to GoyetQ116-1. For the time being, I chose to keep that for later and focus on building this tree.


The next thing I wanted to do was to see how Ganj_Dareh_N fit into all this and if anything beyond Anatolia, Natufians or IBM, and Tianyuan was necessary. To a bit of my surprise, input from Iberomaurusians was quite low, and an extra edge from Mota was needed. Could there be another expansion via a southern route to the Persian Gulf? Potentially, or having more samples near the time of Dzudzuana would show a more deeply branched group in the region. Many things are possible here. While that extra input is small, the Z-score warranted the extra branch to keep that Z-score as low as it is.


The next thing I wanted to check on was how this would change with GoyetQ116-1 being inserted for Kostenki14. I think the results are a confirmation of sorts for some introgression from a Dzudzuana-related pop into all UP Europeans, except for GoyetQ116-1. Meaning that Mota-related input does increase across the board.


These results are pretty interesting and show that using a branch that splits from a line to Mota does work as a source of deeper ancestry in West Eurasians. The potential artifact with African samples, along with more samples being needed to see the formation of the population Mota comes from are definitely needed to see if the same branching can hold, or if Mota is admixed with some type of Eurasian himself. Maybe Mota is mostly derived from a branch that separated from “Bssal Eurasian” well before admixing into the Near East and North Africa. The interesting thing is that even without Dzudzuana, the results still show Iran needing more input from the deep branch of ancestry and also sizeable ENA input. This is an interesting and exciting time. Hopefully, we will get more samples from Africa and West Asia throughout the Late Pleistocene to see what was going on there.




Feldman et al (2019) Late Pleistocene human genome suggests a local origin for the first farmers of Anatolia.

Lazaridis et al (2016) Genomic insights into the origin of farming in the ancient Near East.

Lazaridis et al (2018) Paleolithic DNA from the Caucasus reveals core of West Eurasian ancestry.

van de Loosdrecht et al (2018) Pleistocene North African genomes link Near Eastern and sub-Saharan African human populations.


Steppe Maykop

Thanks to Wang et al (2018), we got our first look at a group referred to as “Steppe Maykop”. Chernykh (2008) does briefly describe this group in his work, “Formation of the Eurasian “Steppe Belt” of stockbreeding cultures: viewed through the prism of archaeometallurgy and radiocarbon dating”.  He goes on to explain that this is a kurgan using culture outside of the Maykop proper zone,  north of the Kuban and Terek Basins, between the Sea of Azov and Caspian Sea (Chernykh, 2008). These steppe complexes contained Maykop cultural items (mostly pottery) (Chernykh, 2008). The image below, taken from Chernykh (2008), illustrates the location of Steppe Maykop.


Fig. 4.
Circumpontic Metallurgical Province at the early formation stage (proto-CMP).
1 – Maykop culture proper;
2 – sites of the so-called Maykop type (“Steppe Maykop”);
3 – Kura-Araks culture;
4 – late (northern) sites of the Uruk community
DNA results of Steppe Maykop
In the supplementary information of Wang et al (2018), the Steppe Maykop samples showed T2e, H2a1, and U7b. The one Y-DNA resulted in Q1a2. These are some interesting results, especially U7b and Q1a2.
U7b is likely rooted in West Asia between 5-10 KYA (Sahakyan et al, 2017), as it is mostly found in West Asia and Southeast Europe. It almost completely absent from Central Asia, where U7 and U7a are more typically found (Sahakyan et al, 2017). Ancient remains from Central Asia are also giving the same story, with U7 and U7a being the only ones found (Narasimhan et al, 2018). Also interesting is the hot spot of U7b in Georgia (Sahakyan et al, 2017), but this is based on modern populations. These Steppe Maykop samples are our first ancient samples with U7b. With their connection to Maykop, that may be the likely source of this haplogroup.
Q1a2 is a common haplogroup of Northern Asia and the Americas, showing that relationship to Native Americans that was discussed in Wang et al (2018). Today, it is fairly common among Turkic speakers of Siberia and Northern Central Asia, along with groups in the Americas.
Modeling with qpGraph
I started first by building a tree that included Chimp, Ust_Ishim, Botai, West_Siberia_N, Maykop_Novosvobodnaya, and Progress_Eneolithic. These samples come from Narasimhan et al (2018), Jeong et al (2018), and Wang et al (2018).
The next thing I did was add Steppe_Maykop to the column designating labels, but did not manually add them to any tree, to find the most significant scores for the first placement. The worst f-stat was Botai, Steppe_Maykop; West_Siberia_N Steppe_Maykop (Z>-49).
To determine if this stat was driven more by Botai or West_Siberia_N, I looked at the Z-scores where only the two were used (Botai, Steppe_Maykop; Botai, Steppe_Maykop and West_Siberia_N, Steppe_Maykop; West_Siberia_N, Steppe_Maykop).
The highest Z-score between the two was -42 between Botai and Steppe_Maykop, versus -30 for West_Siberia_N.
Trying to model this admixture as coming from just West_Siberia_N and not Botai left many 0 drift edges and a worst f-stat approaching 4, likely seeking admixture from Botai and more from Maykop, making it look less feasible. See first graph below. Second one below is the one where Steppe Maykop first branches off of Botai.
This graph left a worst f-stat asking for an edge from Progress_Eneolithic to Steppe_Maykop.
This last graph did have the Steppe_Maykop requiring more Maykop admixture than Progress_Eneolithic, more in-line with the f3-ratio stats (see below). However, I wanted to look at this another way as another way of looking at this created worst Z-scores asking for admixture from EHG, possibly pointing to this group have steppe ancestry that is more like Khvalynsk. Resulting in the tree below.
I also looked at this admixture potentially coming from Early_Maykop, as the f-3 score for both Maykop groups was similar. This also worked well.

This last one was more in-line with f3-ratio outputs and also resolved the need for extra EHG when removing the 0 drift edge in there.


I decided to look at f3-ratio as well, to look for confirmation of these results. The results showed significant scores where the two sources were like Botai and Early_Maykop/Maykop_Novosvobodnaya. So, there is some good confirmation here that the models line up.
Source 1 Source2 Target f_3 std. Error Z SNPs
Botai E_Maykop St_Maykop -0.008599 0.002107 -4.081 68545
Botai Maykop_N St_Maykop -0.008444 0.001574 -5.364 126581
Botai Progress St_Maykop -0.003914 0.001447 -2.705 141243
Botai Khvalynsk St_Maykop 0.005247 0.001706 3.075 106850

The f3-ratio using West_Siberia_N as the source 1 pop did produce more significant scores than Botai, but this could be due to the fact that Siberia_N is rather intermediary to Botai and Progress, as well as Khvalynsk.



When looking back at all the data, combining the uni-parental markers, along with the statistical output, it seems that a population from around the Urals did move into the region north of Maykop proper, and acquired their cultural material (Chernykh, 2008), along with their genetic input. Since Maykop and Steppe Maykop both have the same carbon dates, it could be that Early Maykop is the better source, but there is limited samples here across that time-frame to really come to that conclusion. Progress and Khvalynsk, or something in-between make a good fit as the steppe ancestry in these ssamples. All-in-all it seems very possible for this to be the story of the genesis of Steppe Maykop, from a source more related to Native Americans, along with more typical West Asian ancestry.



Chernyk (2008) Formation of the Eurasian “Steppe Belt” of stock-breeding cultures.

Jeong et al (2018) Characterizing the genetic history of admixture across inner Eurasia.

Narasimhan et al (2018) The genetic formation of South and Central Asia.

Sahakyan et al (2017) Origin and spread of human mitochondrial DNA haplogroup U7.

Wang et al (2018) The genetic history of the Greater Caucasus.