search this blog

Thursday, September 17, 2015

Uralian genes

Below is a graph based on a couple of D-statistics. Note that out of the five Uralic-speaking groups, only Hungarians remain under the upward-sloping red line, clustering among Indo-European-speakers. On the other hand, the Indo-European-speaking Russians from Kargopol cluster with Uralic speakers. However, Russia is a complex affair in this context, because much of the north and east of European Russia was Uralic-speaking until very recently. The relevant D-stats are available here.

See also...

Ancient genomes from NE Europe suggest the tandem spread of Siberian admixture and Uralic languages into the region >3,500 ya


Romulus the I2a L233+ Proto Balto-Slav, layer of Corded Ware Women said...

Interesting, I think EHG would cluster with Uralic though.

I would find it very interesting, and would donate some cash to your blog if you calculated an Fst score of the ancient samples and a Neanderthal group. I've always wondered which group carries the most Neanderthal.

Davidski said...

EHG doesn't look very Uralic on this graph. Uralic languages came from the east after EHG.

I'm not quite sure how to estimate Neanderthal admixture with the tools I have. Fst won't be very informative. I'd need to use something like f4 ratios. I'll have a think about it.

Shaikorth said...

Nganasans and all Samoyedic speakers for their part got their language relatively recently from the southwest when Uralic spread from the Volga-Kama region. That ancient DNA is lacking in both the proto-Uralic zone and pre-Uralic Neolithic and Bronze Age North Russia is a problem. What does the graph look like with Nganasan replaced with Kets, Karitiana and, depending if they have markers, Karasuk/Okunevo?

Davidski said...

None of these work as well as the Nganasan.

Shaikorth said...

Overall pattern works in every case though, and the thing that differentiates Nganasans from all those four is that they have been uralicized recently, and thus share ancestry with western Uralic groups that postdates Karasuk and Okunevo. A slight difference is expected. I assume Anzick, Saqqaq and Dai would all show a similar picture, as would a bronze age genome from the Volga-Kama area.

Kristiina said...

Shaikorth, I agree with you that the overall pattern is the same in all graphs. In all graphs, excluding Karasuk, 4 Uralic groups (Saami, Finnish, Mordovian, Estonian) + Turkic-speaking Chuvash and Russian speaking Kargopol are above the line and Hungarian remains below the line. The graphs with Nganasan, Ket and Karitiana all work very well.

It is obvious from all admixture runs that Uralic groups have high Siberian ANE, significant traces of Native American ancestry and they share the Siberian component.

Moreover, it should not be surprise that the Siberian component is present in all groups inhabiting or bordering Siberia and that it is present in Uralic groups and that it distinguishes them from European IE groups. IMO, it is highly natural! Already a long ago, I proposed that Uralic languages are a result of a merger between IE-resembling language and Siberian/Native American substrate(s).
According to the Ket paper, for example Finns possess c. 5-6% of Arctic Ket component and that substrate is also found in other Uralic/Turkic groups.

Lithuanians have 40.5% N1c and do not have any Siberian and 2500 BC old N1c has been found in Smolensk, so until we get ancient yDNA from Volga-Ural or Finland, we cannot be sure that yDNA N1c introduced Siberian to Finland.

Notwithstanding this, I guess that N1c came from Volga to the Baltic area, and at that time Siberian was present in Volga area and hence N carried Siberian but it is only part of its ancestry. The identification of yDNA from Oleni Ostrov in Kola peninsula is probably key for the understanding of the excess of Siberian in Saami.

la señora bibiloni said...

Hi David, is it possible to make some aDNA comparison with the Atapuerca ATP2 sample and the Bell Beakers from Quedlinburg (R1B-P312-S116) and Kromsdorf (R1B-M269)? Thanks!

andrew said...

It has, of course, been known for a long time that the Hungarian language was a result of language shift in the historic era to the language of a conquering linguistically Uralic aristocracy that left almost no population genetic impact.

Matt said...

Re: trying to use the Nganassan and Yamnaya stat to generate info about IE expansions, it seems pretty tough to add these stats into the kind of PCA we were doing before.

When put straight into that PCA the Nganassan stat ends up correlating with the EHG / MA1 stats (because it has MA1 ancestry, and both lack Basal Eurasian), while the EN stats end up at the opposite end to Nganasan, then you get Yamnaya and WHG in the middle as Yamnaya is strongly correlated with EHG and MA1 (which correlates to Nganassan), and WHG correlates to EN (because EEF has a lot of WHG clade ancestry). So rather than Nganasan get its own PC it gets split between a PCA that positively loading with EHG/ANE ancestry vs EN and one where it positively loads with WHG vs Yamnaya.

E.g. (

OTOH, you could do a regression between Nganassan and EHG / WHG / MA1 / SHG ancestry and find out what the Uralic+Russian outliers "should" be on Nganassan based on their mesolithic sharing... but since that level of mesolithic stat has been reduced by the introduction of Nganasan like ancestry, that doesn't actually do anything.

So I can't think of anything. I'm not sure there is a way around it to get the "right" D stats for the pre-East Asian admixture ancestors of these populations other than if there were some way to specifically mask samples for East Asian ancestry and then run the D-stat on the unmasked portion.

Btw, have you checked out the new "The Role of Recent Admixture in Forming the Contemporary West Eurasian Genomic Landscape" -

It looks like some of the spurious signals from the previous GLOBETROTTER / finestructure have been sorted out (like East Eurasian signals to the Eastern European populations who don't seem to have it by unlinked methods etc.). See their Fig 2 and 4. These methods are promising for looking at the recent history of admixture, which will be between quite closely related populations for which haplotype data is useful. The estimates seem a little inflated though. Whether it now overall makes sense, I don't know.

Shaikorth said...

Matt, it should be fairly easy to do such estimates using ratio-based methods. For Human Origins set, just look at Haak S9.27.

Kristiina said...

Please, correct me if I am mistaken, but it looks like Eskimoan ancestry (c. 67% ANE), unlike Saqqaq (, Figure 3), is without recent East Asian ancestry, i.e. Siberian component. Eskimoan ancestry has also been detected in Ma1, 24 000 BC, so it should be old. Maybe the East Asian part of Eskimos is in part related to the mythical Paleo-Melanesian ancestry detected in some Native Americans (Aleuts and Amazonians). Instead, Siberian ancestry was recognized in Kennewick man, 7000 BC and in Saqqaq (2000 BC), but not in Anzick, 11000 BC. I presume that the Siberian ancestry has some post-Ice Age South East Asian ancestry in it, and that’s why it is rejected by ancient North Eurasian such as Ma1. I am waiting for an analysis of Ainu and I am particularly interested in their relation to Siberian. It is a pity that in all admixture analyses there are only the Japanese who are autosomally quite close to Chinese.

Davidski said...

la señora bibiloni,

As far as I can see, the R1b Bell Beakers from Allentoft and Haak overall show more affinity to Afanasievo, Corded Ware and Yamnaya than to ATP2.

ATP2 Afanasievo Bell_Beaker_R1b Mbuti -0.0112 -1.158
ATP2 Corded_Ware Bell_Beaker_R1b Mbuti -0.0292 -3.316
ATP2 Yamnaya Bell_Beaker_R1b Mbuti -0.0044 -0.524

I can't run the Kromsdorf Bell Beakers from Lee et al. in this test.

Unknown said...

Newish paper on Alans, Sarmatians etcАфанасьев_Г.Е._Ван_Л._Вень_Ш._Вэй_Л._Добровольская_М.В._Коробов_Д.С._Решетова_И.К._Ли_Х._Тун_С._Хазарские_конфедераты_в_бассейне_Дона_Тезисы_докладов_на_Всероссийской_научной_конференции_Естественнонаучные_методы_исследования_и_парадигма_современной_археологии_._М._ИА_РАН._2015

Davidski said...

Those samples are very late and don't say anything about the early Indo-European or even Indo-Iranian expansions. They mostly look like backflow from Asia just before the Turkic expansion into Europe.

Aram said...

The two aDNA marked Sarmatians are J1. Age: 2nd - 3rd century AD.
They are from Beslan, North Ossetia.

Unknown said...

Indeed, they're Middle Ages

But it does confirm that the steppe was a conduit for population movements.

Unknown said...

The R1a of the Alans might be of the Z93 -derived variety (?)

Davidski said...

It is. It also looks like the same subclade under Z93 as the Sintashta Z93.

Unknown said...

So Z93 moved south then at some point indo-Iranian tribes moved north to the Pontic steppe,
probably Iron age.

They then disappeared from west of the Dnieper in late antiquity c. 400 AD. Hence the gap between european Russians and Caucasians.

Nirjhar007 said...

Interesting points Mike.

Davidski said...

They weren't Indo-Iranian tribes but, more specifically, Iranian tribes like Scythians and Sarmatians from the Asian steppe, even from as far as the Altai. This is where European-derived groups like Andronovo mixed with Asians, including West Asians, who moved up the inner Asian corridor into the steppe, to form new cultures.

This process is discussed in Allentoft et al.

Unknown said...

True, lets call them Scythians, not Indo-Iranians. But they weren't from the Altai.

And it's likely they brought Indo-Iranian languages, specifically East Iranian to the western steppe, although Trubachev even found the odd Indian toponym around the Black Sea.

Nirjhar007 said...

What is that Toponym? I would love to see :).

Unknown said...

I don't know , cant remember. The reference is in J Nichol's chapter "eurasian Spread zone ..." in Archaeology & Language II.

Davidski said...

They were from around the Altai and even south Siberia. You'll see this when their genome-wide data gets published. They'll be similar to some of the Karasuk genomes.

Nirjhar007 said...

Ah I see, Thx.

Unknown said...


But didn't Karasuk have Q and lack G2 ?

The point of origin might still very well have been Central Asia. Indeed, this is what the archaeological and literary data suggests .

Whatever the case, im dying to see full genome data from Catacomb culture- probably CWC like ?

Davidski said...

RISE552 is late Yamnaya/early Catacomb.

Here's a paper about the burial.

Slumbery said...

andrew said...
"It has, of course, been known for a long time that the Hungarian language was a result of language shift in the historic era to the language of a conquering linguistically Uralic aristocracy that left almost no population genetic impact."

It is accepted as a trivia for long time, but not really proven. It was probably more complicated than this simplistic elite domination story.

The genetic study of the remains of early 10th century Hungarian elite are somewhat strange, because on PCA they are far too in the direction of East Asian (a way East from modern Kazakhs) for the supposed source of an Uralic language. Almost as they where a not yet fully mixed Turkic group from Western Mongolia or so...
Everybody talks about this small elite population assimilates the overwhelmingly more numerous commoners like a trivial things, but I yet to see a convincing historical example where this actually happened under similar conditions and with this speed. There are very well known examples of the opposite though. So when we say that a small elite group of Hungarians assimilated the Slavic masses that is actually an extraordinary claim that needs extraordinary evidence.

Unknown said...


I agree .
The Magyars probably numbered thousands entering the Carpathian basin , with thousands of slavs amd proto-Romanians, etc already there.
So certainly a significant Impact.

Slumbery said...


As far as we know the 9th century population of the Carpathian Basin and Transylvania was mostly Slav with an Avar minority that was probably already slavified to an unknown extent (but was still mostly distinguishable in grave artifacts). "Proto-Romanians" are not impossible but not attested. The Slavic presence even in the early 11th century Transylvania is well documented both in toponyms and personal names, etc., but there is absolutely no trace of any Latin population. They arrived from the south in the 12th century (and they are well documented from that time).
Some (population-wise) marginal Bulgarian presence in the South and "Frank" in the west (Mosaburg) is also attested, but that is about all.

Unknown said...


Actually I agree. But the Vlach debate, Gesta, etc is beyond the scope of this blog ha ha

But my main point was that the hunt gerunds themselves might of been rather numerous. And what's more they came from the Black Sea region, and not the Ural mountains. So they would've been already very much Eastern European overall

Slumbery said...


A growing number of modern Hungarian archaeologist say that the time Hungarians spent in the "Black Sea region" was very short. Archaeologically the presence of Hungarians so far only attested from West Ukraine and Moldavia in the middle / second half of 9th century (two or three generations top) and before that from the Ural region from East of Volga Bulgaria.

But I agree with your point. It just that they may very well where totally European genetically even in the SW Ural, setting aside the apparently Turkic part of their elite that of course left its trace on the language and self-identification.

Unknown said...

We don't know how turkic got into early hungarian. All the turkic toponyms in hungary are from the Cuman period (ie post -conquest). But one can easily imagine being in the black sea steppe there's have been much contact with turkic speakers.

And my impression was that scholars look to Magyars as one of the constituents of the Khazar empire, thus much before the 9th century. A distinctly Magyar 'culture ' was only created c. 850 because that is when those •Magyars" who rebelled needed to proclaim themselves through funerary ritual. Ie material culture is actively manipulated to create identity, and is not a passive reflection of an ancient, ineffible ethnicity going back to the Ural Mountains.

Slumbery said...


Yes, that was the scholarly consensus and they still teach that in the schools in Hungary. But it is lately strongly questioned on archaeological base, also noting that contemporary sources are very vague and unreliable on the matter of Kazars vs. Hungarians and not very early either. Also grave goods clearly connected to Hungarians are found in the Ural region (by Russian archaeologists).

The Turkic influence on Hungarian is very apparent and obvious into the extent that the Turkic relatedness of the language was seriously considered against the Uralic theory in the 19th century. Detailed scholarly research showed that the Uralic "core" is older and bigger and that convinced the Turkic "party" of the Academy back then, but there was a reason why the Turkic relatedness came up at all. This is nothing to do with toponyms, Turkic is everywhere in the living language. However the toponyms that have "Turkic" origin in this way are simply counted as Hungarian, because this was already a "digested" part of the Hungarian language in the Carpathian Basin.
Also note that the Turkic part of the Hungarian language came from a Turkic dialect that already went extinct (except maybe the Chuvash), but was spoken by the Bulgarians for example (and presumably the Kazars).
A decent book on this particular topic:

Unknown said...


Very interesting, thanks. Ill have a read

Maybe Porphyrogenitus was not too wrong when he called the Magyars τουρκοι :)

I guess it might bring some reconsideration of previously rejected hypotheses that other languages, like Old Bulgar also had a Uralic base ?

Kristiina said...

@Slumbery ”The genetic study of the remains of early 10th century Hungarian elite are somewhat strange, because on PCA they are far too in the direction of East Asian”

Is this conclusion based on craniological analysis? I am only aware of a paper that conluded that N1c was present among ancient Hungarians (2/7), but unfortunately they did not analyze other yDNA haplogroups (

Do you know what other yDNA’s ancient Hungarians carried?

With reference to this recent Afanasev et altri study, it is interesting to note that the above-mentioned Hungarian paper says that ”in the present study, haplogroup J was unexpectedly common in the Hungarian-speaking populations (Hungarians: 16%, Szeklers: 21.6%) considering that Sarmatian yDNA 100-200 AD in South Russia (Beslan) was J1 and Saltovo-Mayatskaya yDNA 700-800 AD in Podsk was J2a.
Ancient Magyar mtDNA is the following, so it is not particularly eastern:
2xN1a; 1xI; 2xX
1xHV; 1xV; 7xH; 2xT(xT2); 2xT2
1xU; 1xU3; 2xU4; 1xU5a1
(Comparison of Maternal Lineage and Biogeographic Analyses of Ancient and Modern Hungarian Populations:

Turkic speaking Volga Chuvash have a similar mtDNA pool:
1.8% N1a; 1.8% I;
1% HV; 6% V; 27% H; 3.6% T1;
U3 1.8%,; U4 16%; U5 14.5%; U8 1.8%; K 7%
M 1.8%
However, they have also K, J, C, D and A and lack X and B.

Mike, if N1c and N1b came from Volga and originally mostly spoke a Uralic type language, which is of course not proven, there could very well be a Uralic substrate in Turkic as N1c and N1b reach high frequencies in some Turkic groups and their expansion seems rather to correlate with the later Turkic expansion.

Slumbery said...


I was referring to Fig.4 (biogeographic abalyses if Hungarian populations) of the article you mention (Comparison of Maternal Lineage and Biogeographic Analyses of Ancient and Modern Hungarian Populations.)

Look at the position of the population called "classicals". They are totally off and notably nowhere close to any included modern Uralic or Siberian population.
The same is true about any modern Turkic tough. I remembered it differently after a few years.

By the way, the "ancient" in the article is a sum of "classical" + "commoner" and the authors claim only the "classicals" were Hungarians.

And I do not know any other study of eraly Hungarian YDNA than the one you also know.

Kristiina said...

It is true that Classicals are totally off, but the position of Chuvash could be somewhere near. Also Uralic Maris could not be that far. Among Uralic groups, Saami and Finns are not really a good reference, as they both have a lot of Fennoscandian/Baltic ancestry and the Saami in particular have a significant portion of Arctic ancestry.

I did not think about that division between classical and commoner, but it looks like the mtDNA haplogroups I gave include both groups. Classicals have 2xH, 2xN1a, 2xU4 and 1xX, 1xB, 1xR, 1xT and 1xU. This fullfils my expectations of a proto-Hungarian mtDNA pool although this is surely just a small part of it.

Unknown said...

How am I supposed to run this?

Davidski said...

It's not an ancestry tool, just a demonstration.