A comedy of errors

A couple of years ago, the authors of a paper about a group of Iron Age nomads from the site of Shirenzigou, in the eastern Tian Shan, made a mistake. They wrongly assigned two of these nomads to Y-haplogroup R1b-M269.

This faux pas made them believe that the Shirenzigou nomads were closely related to the M269-rich population associated with the Afanasievo culture.

Indeed, since the Afanasievo culture was often credited with the spread of Tocharian languages to the Tarim Basin, these authors, led by Chao Ning, also concluded that the Shirenzigou nomads were potentially the missing link between the Afanasievo culture and the Tocharians (see here).

Moreover, Ning et al. used formal statistics to argue that the Shirenzegou nomads harbored Afanasievo-related genome-wide ancestry, rather than Sintashta-related genome-wide ancestry, despite the fact that the latter ancestry was widespread in the Tian Shan and surrounds during the Bronze and Iron ages. Soon after, another group of authors, led by Chuan-Chao Wang, also went out of their way to link the Shirenzigou nomads to the Afanasievo people with genome-wide DNA using formal statistics (see here).

Interestingly, one of the Shirenzigou nomads belongs to Y-haplogroup R1a-Z93, which is an obvious Sintashta-related lineage. Both Ning et al. and Wang et al. missed this important fact.

They also missed the key fact that the R1b lineage found in the Shirenzigou nomads actually belongs to an Inner Asian subclade, which is only very distantly related to the originally Eastern European R1b-M269.

Now, formal stats are a very useful tool for studying genome-wide ancestry. But they're not infallible, and that's actually something of an understatement. Indeed, if you don't run sanity checks when using formal stats, you're likely to come to some unusual, even arse about face, conclusions. Uniparental markers, like Y-chromosome haplogroups, can provide a robust sanity check when running formal stats on genome-wide data.

One problem with formal stats is that Sintashta-related ancestry often looks very much like Afanasievo-related ancestry when it's mixed with indigenous Central Asian ancestry. Basically, the reason why this happens is that the Central Asian ancestry dampens the Early European Farmer (EEF) signal in the Sintashta-related ancestry.

This is an artifact that once caused scientists at Harvard to believe that Central Asian Scythians and present-day South Asians lacked Sintashta-related ancestry.

Unfortunately, since the publication of the Ning et al. paper, a consensus has emerged in academia that the Shirenzigou nomads are indeed the missing link between the Afanasievo culture and the Tocharians. But, let's be objective and honest here, it's a consensus based on nothing more than a comedy of errors.

On the other hand, me and most of the commentators at this blog have formed opinions about the Shirenzigou nomads that are totally at odds with the academic consensus, that:

- they're a complex mixture of Sintashta-related, indigenous Central Asian and Tibetan-related ancestries, with no clear, unambiguous signal of Afanasievo-related ancestry

- they weren't the speakers of Proto-Tocharian or even related in any specific way to the Tocharians

- they were probably the speakers of a now extinct Indo-Iranian language, and, at least based on geographic proximity, possibly related to the Yuezhi.

Feel free to make up your own mind. But for me, the question of how Tocharian languages ended up in the Tarim Basin remains wide open. I admit though, I'm currently quite partial to the idea floated here by commentator Copper Axe that the Chemurchek culture may have had something to do with it.

See also...

Don't believe everything you read in peer reviewed papers

The origins of East Asians (Wang et al. 2020 preprint)

Over at bioRxiv at this LINK. Here's the abstract:

The deep population history of East Asia remains poorly understood due to a lack of ancient DNA data and sparse sampling of present-day people. We report genome-wide data from 191 individuals from Mongolia, northern China, Taiwan, the Amur River Basin and Japan dating to 6000 BCE - 1000 CE, many from contexts never previously analyzed with ancient DNA. We also report 383 present-day individuals from 46 groups mostly from the Tibetan Plateau and southern China. We document how 6000-3600 BCE people of Mongolia and the Amur River Basin were from populations that expanded over Northeast Asia, likely dispersing the ancestors of Mongolic and Tungusic languages. In a time transect of 89 Mongolians, we reveal how Yamnaya steppe pastoralist spread from the west by 3300-2900 BCE in association with the Afanasievo culture, although we also document a boy buried in an Afanasievo barrow with ancestry entirely from local Mongolian hunter-gatherers, representing a unique case of someone of entirely non-Yamnaya ancestry interred in this way. The second spread of Yamnaya-derived ancestry came via groups that harbored about a third of their ancestry from European farmers, which nearly completely displaced unmixed Yamnaya-related lineages in Mongolia in the second millennium BCE, but did not replace Afanasievo lineages in western China where Afanasievo ancestry persisted, plausibly acting as the source of the early-splitting Tocharian branch of Indo-European languages. Analyzing 20 Yellow River Basin farmers dating to ~3000 BCE, we document a population that was a plausible vector for the spread of Sino-Tibetan languages both to the Tibetan Plateau and to the central plain where they mixed with southern agriculturalists to form the ancestors of Han Chinese. We show that the individuals in a time transect of 52 ancient Taiwan individuals spanning at least 1400 BCE to 600 CE were consistent with being nearly direct descendants of Yangtze Valley first farmers who likely spread Austronesian, Tai-Kadai and Austroasiatic languages across Southeast and South Asia and mixing with the people they encountered, contributing to a four-fold reduction of genetic differentiation during the emergence of complex societies. We finally report data from Jomon hunter-gatherers from Japan who harbored one of the earliest splitting branches of East Eurasian variation, and show an affinity among Jomon, Amur River Basin, ancient Taiwan, and Austronesian-speakers, as expected for ancestry if they all had contributions from a Late Pleistocene coastal route migration to East Asia.

Also this part is interesting, but surprisingly naive:

The findings of the original study that reported evidence that the Afanasievo spread was the source of Steppe ancestry in the Iron Age Shirenzigou have been questioned with the proposal of alternative models that use ancient Kazakh Steppe Herders from the site of Botai, Wusun, Saka and ancient Tibetans from the site of Mebrak 15 in present-day Nepal as major sources for Steppe and East Asian-related ancestry [28]. However, when we fit these models with Russia_Afanasievo and Mongolian_East_N added to the outgroups, the proposed models are rejected (P-values between 10 -7 and 10 -2), except in a model involving a single low coverage Saka individual from Kazakhstan as a source (P=0.17, likely reflecting the limited power to reject models with this low coverage). Repeating the modeling using other ancient Nepalese with very similar genetic ancestry to that in Mebrak results in uniformly poor fits (Online Table 5). Thus, ancestry typical of the Afanasievo culture and Mongolian Neolithic contributed to the Shirenzigou individuals, supporting the theory that the Tocharian languages of the Tarim Basin—from the second-oldest-known branch of the Indo-European language family—spread eastward through the migration of Yamnaya steppe pastoralists to the Altai Mountains and Mongolia in the guise of the Afansievo culture, from where they spread further to Xinjiang [5,7,8,27,29,30]. These results are significant for theories of Indo-European language diversification, as they increase the evidence in favor of the hypothesis the branch time of the second-oldest branch in the Indo-European language tree occurred at the end of the fourth millennium BCE [27,29,30].

I'd say the authors are putting too much faith in their qpAdm mixture models. They ought to know that qpAdm has some serious limitations, especially in regards to fine scale ancestry. I would urge them to become better acquainted with the uniparental markers of the Iron Age Shirenzigou samples instead of forcing the ideas that these individuals harbor Afanasievo-derived ancestry and lack Tibetan-related ancestry.

See also...

They mixed up Huns with Tocharians

A surprising twist to the Shirenzigou nomads story

Afanasievo people may well have been proto-Tocharian speakers (Ning et al. 2019)

A surprising twist to the Shirenzigou nomads story

Remember those potentially Afanasievo-derived and Tocharian-related Shirenzigou nomads from the Ning et al. paper? Well, in my opinion, they're probably neither. The genotypes and other data for these Iron Age individuals from the eastern Tian Shan are available here.

Below are a few successful and not so successful qpAdm mixture models for them. Note that I tried to use a wide range of relevant "right pops", but also retain a lot of markers, specifically to be able to discriminate between different types of steppe and steppe-derived sources of gene flow (refer to the full output). Admittedly, the Shirenzigou nomads can be modeled with Afanasievo-related ancestry, but...

CHN_Shirenzigou_IA
KAZ_Botai 0.161±0.023
KAZ_Wusun 0.490±0.023
NPL_Mebrak_2125BP 0.349±0.019
chisq 5.793
tail prob 0.926172
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.143±0.022
NPL_Mebrak_2125BP 0.295±0.019
Saka_Tian_Shan 0.562±0.024
chisq 6.796
tail prob 0.870794
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.185±0.023
NPL_Mebrak_2125BP 0.428±0.021
RUS_Sintashta_MLBA 0.270±0.026
TJK_Sarazm_En 0.117±0.027
chisq 11.351
tail prob 0.414345
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.032±0.027
KAZ_Zevakinskiy_LBA 0.567±0.025
NPL_Mebrak_2125BP 0.401±0.019
chisq 15.157
tail prob 0.232961
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.452±0.031
RUS_Afanasievo 0.435±0.025
RUS_Okunevo_BA 0.114±0.049
chisq 19.808
tail prob 0.0708003
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.409±0.031
RUS_Okunevo_BA 0.173±0.050
Yamnaya_RUS_Caucasus 0.418±0.026
chisq 20.453
tail prob 0.0589872
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.464±0.033
RUS_Okunevo_BA 0.104±0.053
Yamnaya_RUS_Samara 0.432±0.027
chisq 27.189
tail prob 0.0072566
Full output

Both the Wusun and Saka are generally accepted to have been the speakers of Indo-Iranian languages. So it's possible that the Shirenzigou nomads were Indo-Iranian speakers too, or at least derived from such peoples.

Surprisingly, NPL_Mebrak_2125BP was the key to obtaining the best statistical fits. This is a trio of samples, roughly contemporaneous with the Shirenzigou nomads, from a burial site high up in the Himalayas in what is now Nepal (see here).

To be honest, I'm not quite sure why the Himalayan ancients work so well in my models. Perhaps they're just a really good proxy for an Iron Age population from the northern edge of the Tibetan Plateau?

By the way, most of the Shirenzigou nomads made it into the latest Global25 datasheets (see here). They can be analyzed in a variety of ways described in this blog post: Getting the most out of the Global25. Below is a screen cap of me comparing the effectiveness of Afanasievo, Sintashta and Wusun samples as proxies for the steppe ancestry in the Shirenzigou nomads with an online tool freely available here. As expected, the algorithm picks Sintashta ahead of Afanasievo, and the Wusun ahead of both.

See also...

They mixed up Huns with Tocharians

Some myths die hard

The mystery of the Sintashta people

They mixed up Huns with Tocharians

I don't yet have the genomes from the recent Ning et al. paper on the Iron Age nomads from the Shirenzigou site in the eastern Tian Shan. But I do have most of the previously published data featured in the paper, including the Damgaard et al. 2018 Hun and Saka samples from the western Tian Shan.

After reading the Ning et al. paper between the lines and running a few analyses of my own, it's clear to me that most of the supposedly Tocharian-related Shirenzigou individuals actually share a very close relationship with the Tian Shan Huns, and indeed may have been their ancestors.

For instance, Ning et al. found that a large part of the ancestry of the Shirenzigou ancients could be modeled with the Tian Shan Huns, which was an anachronistic approach because the former are older than the latter. They also found that Ulchi-related ancestry was a key part of the genetic structure of eight out of the ten Shirenzigou individuals, and this likewise appears to be an important part of the genetic structure of the Tian Shan Huns.

Note the strong statistical fits in the Global25/nMonte and qpAdm mixture models below, respectively, which characterize these Huns as a two-way mixture between the Ulchi and the earlier Tian Shan Saka. And keep in mind that the Saka also harbor significant Ulchi-related ancestry.

Hun_Tian_Shan
Saka_Tian_Shan,92
Ulchi,8
distance%=1.2553

Hun_Tian_Shan
Saka_Tian_Shan 0.928±0.009
Ulchi 0.072±0.009
chisq 4.409
tail prob 0.992464
Full output

Moreover, the Shirenzigou males belong to Y-haplogroups Q1a and R1b (two instances of each), and they share the latter with one of the Tian Shan Huns. Judging by the data from the relevant BAM files, it's also possible that the Shirenzigou males share a very rare subclade of R1b with the Hun, defined by the PH155 mutation (see here). The Y-haplogroup assignments for the other Tian Shan Huns end at R and R1, but that's almost certainly due to missing data.

On the other hand, two Tian Shan Sakas belong to Y-haplogroup R1a but none to R1b, which fits with the pattern from currently available ancient DNA that R1a was more common than R1b in Saka-related groups, such as the Scythians and Sarmatians (see here).

This is all very interesting, because the Huns replaced the Saka in the western Tian Shan, and, considering their R1b and excess Ulchi-related ancestry, very likely moved into the region from the direction of Shirenzigou. Indeed, in my opinion a strong argument can now be made that the Iron Age population from the Shirenzigou region took part in the formation of the Hunnic confederacy.

So where does that leave the theory presented by Ning et al. that the Shirenzigou ancients may have been closely related, and perhaps even ancestral, to the Tocharians, simply because they packed a lot of Yamnaya-related and possibly proto-Tocharian Afanasievo ancestry, and were living close to the Tarim Basin, where Tocharian languages were subsequently first attested?

I'm not sure, but I now find it difficult to reconcile this theory with the fact that they were closely related, and probably ancestral, to the Tian Shan Huns. As far as I'm aware, Huns cannot be linked to Tocharians in any meaningful way.

Of course it's possible that different Afanasievo-derived groups were living in the Tarim Basin and surrounds, and, as some merged with new populations pushing into the region from the east and adopted non-Indo-European languages, others retained their Tocharian speech and eventually split into communities speaking Tocharian A, B and apparently also C (see here).

But this has to be demonstrated directly with ancient DNA from archeological sites where Tocharian languages were attested. Till then, I'll keep thinking that Ning et al. wrote a paper about Tocharians that really should've been a paper about Huns.

Here's a famous wall painting of Tocharian princes from the cave of the sixteen sword-bearers in the Tarim Basin, dated to 432–538 AD. They don't look like guys with a lot of Ulchi-related admixture to me, but I might be wrong. Feel free to let me know what you think in the comments below.

Update 08/17/2019: The Shirenzigou nomads are now in my dataset. Below are a few successful and not so successful qpAdm mixture models for them. Note that I tried to use a wide range of relevant "right pops", but also retain a lot of markers, specifically to be able to discriminate between different types of steppe and steppe-derived sources of gene flow (refer to the full output). Admittedly, the Shirenzigou nomads can be modeled with Afanasievo-related ancestry, but...

CHN_Shirenzigou_IA
KAZ_Botai 0.161±0.023
KAZ_Wusun 0.490±0.023
NPL_Mebrak_2125BP 0.349±0.019
chisq 5.793
tail prob 0.926172
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.143±0.022
NPL_Mebrak_2125BP 0.295±0.019
Saka_Tian_Shan 0.562±0.024
chisq 6.796
tail prob 0.870794
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.185±0.023
NPL_Mebrak_2125BP 0.428±0.021
RUS_Sintashta_MLBA 0.270±0.026
TJK_Sarazm_En 0.117±0.027
chisq 11.351
tail prob 0.414345
Full output

CHN_Shirenzigou_IA
KAZ_Botai 0.032±0.027
KAZ_Zevakinskiy_LBA 0.567±0.025
NPL_Mebrak_2125BP 0.401±0.019
chisq 15.157
tail prob 0.232961
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.452±0.031
RUS_Afanasievo 0.435±0.025
RUS_Okunevo_BA 0.114±0.049
chisq 19.808
tail prob 0.0708003
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.409±0.031
RUS_Okunevo_BA 0.173±0.050
Yamnaya_RUS_Caucasus 0.418±0.026
chisq 20.453
tail prob 0.0589872
Full output

CHN_Shirenzigou_IA
NPL_Mebrak_2125BP 0.464±0.033
RUS_Okunevo_BA 0.104±0.053
Yamnaya_RUS_Samara 0.432±0.027
chisq 27.189
tail prob 0.0072566
Full output

Both the Wusun and Saka are generally accepted to have been the speakers of Indo-Iranian languages. So it's possible that the Shirenzigou nomads were Indo-Iranian speakers too, or at least derived from such peoples.

Surprisingly, NPL_Mebrak_2125BP was the key to obtaining the best statistical fits. This is a trio of samples, roughly contemporaneous with the Shirenzigou nomads, from a burial site high up in the Himalayas in what is now Nepal (see here).

To be honest, I'm not quite sure why the Himalayan ancients work so well in my models. Perhaps they're just a really good proxy for an Iron Age population from the northern part of the Tibetan Plateau? By the way, most of the Shirenzigou nomads made it into the latest Global25 datasheets (see here).

See also...

Almost everything you ever wanted to know about the Xiaohe-Gumugou cemeteries

The mystery of the Sintashta people

Late PIE ground zero now obvious; location of PIE homeland still uncertain, but...

Afanasievo people may well have been proto-Tocharian speakers (Ning et al. 2019)

Update 17/08/2019: A surprising twist to the Shirenzigou nomads story

...

During the Early Bronze Age, around 2,900 BCE, a population associated with the Yamnaya archeological culture migrated from the Pontic-Caspian steppe in Eastern Europe deep into Asia, as far as the Minusinsk Basin in South Siberia.

This rapid, long-range expansion was likely to have been the first significant migration of a Yamnaya-related group far to the east of the Ural Mountains, and it resulted in the formation of the Afanasievo archeological culture (see here).

The appearance of Tocharian languages in the Tarim Basin, in what is now western China, is often associated with the Afanasievo culture, mainly because of the confirmed presence of European-related populations in the Tarim Basin during the Bronze Age, as well as the likely highly divergent position of the Tocharian node in the Indo-European language phylogeny.

But the Afanasievo people were separated by considerable distance in space and time from the Tocharians, and can't yet be reliably linked to them with archeological or genetic data. So even though the inference that the former are linguistically ancestral to the latter is quite plausible, it's far from certain.

However, thanks to a new paper at Current Biology by Ning et al., at least we now know that a population with significant Yamnaya/Afanasievo-related ancestry was living in the eastern Tian Shan Mountains just a few hundred years before Tocharian languages were attested nearby [LINK]. Below is the paper summary, emphasis is mine:

Recent studies of early Bronze Age human genomes revealed a massive population expansion by individuals-related to the Yamnaya culture, from the Pontic Caspian steppe into Western and Eastern Eurasia, likely accompanied by the spread of Indo-European languages [1, 2, 3, 4, 5]. The south eastern extent of this migration is currently not known. Modern-day human populations from the Xinjiang region in northwestern China show a complex population history, with genetic links to both Eastern and Western Eurasia [6, 7, 8, 9, 10]. However, due to the lack of ancient genomic data, it remains unclear which source populations contributed to the Xinjiang population and what was the timing and the number of admixture events. Here, we report the first genome-wide data of 10 ancient individuals from northeastern Xinjiang. They are dated to around 2,200 years ago and were found at the Iron Age Shirenzigou site. We find them to be already genetically admixed between Eastern and Western Eurasians. We also find that the majority of the East Eurasian ancestry in the Shirenzigou individuals is-related to northeastern Asian populations, while the West Eurasian ancestry is best presented by ∼20% to 80% Yamnaya-like ancestry. Our data thus suggest a Western Eurasian steppe origin for at least part of the ancient Xinjiang population. Our findings furthermore support a Yamnaya-related origin for the now extinct Tocharian languages in the Tarim Basin, in southern Xinjiang.

Ning et al., Ancient Genomes Reveal Yamnaya-Related Ancestry and a Potential Source of Indo-European Speakers in Iron Age Tianshan, Current Biology, July 25, 2019, DOI: https://doi.org/10.1016/j.cub.2019.06.044

See also...

It was always going to be this way

The mystery of the Sintashta people

Late PIE ground zero now obvious; location of PIE homeland still uncertain, but...