Investigating East Asian admixture in Balkans/Anatolia/Caucasus

I used ALDER with a dataset of populations from the Balkans, Anatolia, and Caucasus, using the She, Japanese, Miaozu, and Dai as East Asian references. A few caveats for this analysis:

1. Some populations may possess "South Asian" admixture which may be mistaken for East Asian
2. Populations differ in the number of SNPs used in the analysis; for example, the Armenian_D sample includes mostly Family Finder data which has a smaller overlap with the SNP set used
3. Populations differ in the number of individuals used, from a low of 5 (e.g., Turkish_Cypriot) to a high of 45 (Armenian_D)

I have also added the HGDP Europeans to the analysis. The results can be seen below; I have made bold those rows where all estimates are at least one standard error above zero, and bold/big those where the estimates are they are two standard errors above zero. I consider the latter to be the most reliable.

I have already discussed the Turkish signal of admixture at length elsewhere. I will note that the Iranian_D sample produces similar or younger admixture dates, which would make sense, given the fact that the Iranians came under control of the Mongols, while the successes of the latter in Anatolia were short-lived.

A very interesting signal is that of the North Ossetians which show admixture ~9-10 centuries ago. This seems to have occurred a little after the foundation of the kingdom of Alania, and I think it makes excellent sense to view it as the signal of Eurasian nomads (who must have carried some East Asian admixture at that time) intermingling with pre-Iranic local Caucasus populations, Two other populations from the Caucasus, the Georgians and Lezgins (and also the Abkhaz and Chechens) show earlier admixture signals that could very well date to the period of east-west Eurasian migrations inaugurated by the Huns, although a possible Sassanian origin of such influence cannot be overlooked.

The Kurds are another interesting case where the Dodecad sample and the Yunusbayev et al  sample produce very different dates. The different number of SNPs may be at play here, or it may be that some Dodecad participants have recent Turkish ancestry that cause the admixture date average to appear lower, although the globe13 analysis suggests that the "East Asian" found here may be in fact "South Asian".

It seems to me that with large, dense, and well-curated sample sets from several of these populations, their admixture dynamics will become more distinct.

Huge study on Y-chromosome variation in Iran (Grugni et al. 2012)

This is the equivalent of a box of candy for anyone interested in Eurasian (pre-)history. I will have digest all the goodies within, and post any of my comments as updates to this post.

UPDATE I: Here is the table of haplogroup frequencies for easy reference:

One of the most interesting finds is the presence of a few IJ-M429* chromosomes  in the sample. Haplogroup IJ encompasses the major European I subclade, and the major West Asian J subclade. The discovery of IJ* chromosomes is consistent with the origin of this haplogroup in West Asia; it is widely believed that haplogroup I represents a pre-Neolithic lineage in Europe, although at present there are no Y chromosome-tested pre-Neolithic remains.

There is also a wide assortment of Q and R in Iran. While some of these may be intrusive (e.g., the 42.6% of Q1a2 in Turkmen, likely a legacy of their Central Asian origins), the overall picture appears consistent with a deep presence of these lineages in Iran. This is especially true for haplogroup R where pretty much every paragroup and derived group is present, excepting those likely to have originated recently elsewhere.

UPDATE II: From the paper:

Although accounting only for 25% of the total variance, the first two components (Figure 3) separate populations according to their geographic and ethnic origin and define five main clusters: East-African, North-African and Near Eastern Arab, European, Near Eastern and South Asian. The 1stPC clearly distinguishes the East African groups (showing a high frequency of haplogroup E) from all the others which distribute longitudinally along the axis with a wide overlapping between European and Arab peoples and between Near Eastern and South Asian groups. The 2ndPC separates the North-African and Near Eastern Arabs (characterized by the highest frequency of haplogroup J1) from Europeans (characterized by haplogroups I, R1a and R1b) and the Near Easterners from the South Asians (due to the distribution of haplogroups G, R2 and L). Iranian groups do not cluster all together, occupying intermediate positions among Arab, Near Eastern and Asian clusters. In this scenario, it is worth of noticing the position of three Iranian groups: (i) Khuzestan Arabs (KHU-Ar) who, despite their Arabic origin, are close to the Iranian samples; (ii) Armenians from Tehran (THE-Ar), whose position, in the upper part of the Iranian distribution, indicates a close affinity with the Near Eastern cluster, while their position near Turkey and Caucasus groups, due to the high frequency R1b-M269 and other European markers (eg: I-M170), is in agreement with their Armenia origin; (iii) Sistan Baluchestan (SB-Ba) that clusters with its neighbouring Pakistan.

UPDATE III: There are lots of little details in the haplogroup distribution that make historical sense. For example, C3 exists in Assyrians from Azarbaijan, and both C*, C3, and O exists in Zoroastrians from Yazd. It is often forgotten that before the spread of Islam, and quite time thereafter, Inner Asia was teeming with Zoroastrians and Nestorian Christians. It seems quite likely that these outliers represent a legacy of these communities.

UPDATE IV: I have a feeling that Razib will take exception with this statement: "Ancient Persian people were firstly characterized by the Zoroastrianism. After the Islamization, Shi'a became the main doctrine of all Iranian people."


UPDATE V: This confirms my observation from the recent studies in Afghanistan, that there is an inverse relationship of J2a and R1a in Iranian-speaking groups, with an excess of the latter among the eastern Iranians, and of the former among the Persians. From the paper:

Among the different J2a haplogroups, J2a-M530 [46] is the most informative as for ancient dispersal events from the Iranian region. This lineage probably originated in Iran where it displays its highest frequency and variance in Yazd and Mazandaran (Figure 2). Taking into account its microsatellite variation and age estimates along its distribution area (Tables S3 and S7), it is likely that its diffusion could have been triggered by the Euroasiatic climatic amelioration after the Last Glacial Maximum and later increased by agriculture spread from Turkey and Caucasus towards southern Europe. The high variance observed in the Italian Peninsula is probably the result of stratifications of subsequent migrations and/or of the presence of sub-lineages not yet identified. Of interest in the M530 network (Figures 2 and S3) is the presence of a lateral branch that is characterized by a DYS391 repeat number equal to 9. Differently from previous observations [46], this branch is not restricted to Anatolian Greek samples being shared with different eastern Mediterranean coastal populations. The M530 diffusion pattern seems to be also shared by the paragroups J2a-M410* and J2a-PAGE55*. In addition, the variance distribution of the rare R1b-M269* Y chromosomes, displaying decreasing values from Iran, Anatolia and the western Black Sea coastal region, is also suggestive of a westward diffusion from the Iranian plateau, although more complex scenarios can be still envisioned because of its non-star like structure.

Of course, the idea that the diffusion of J2a related lineages ties in with early agricultural expansions has been with us for a long time, but it is time to abandon it. First of all, as we have seen, J2a diminishes greatly as we head towards South Asia; it certainly doesn't look like the lineage of the multitude of agricultural settlements that sprang up along the southeastern vector soon after the invention of agriculture. Second, it is lacking so far in all ancient Y chromosome data from Europe down to 5,000 years ago. It seems much more probably that J2 related lineages spread from the highlands of West Asia much later. 


The "age estimates" are the result of using the inappropriate "evolutionary mutation rate", and become even older because of the inclusion of the DYS388 marker that is very stable in many haplogroups but very mutable within haplogroup J. On the left you can see frequency, Y-STR variance, and haplotype network structures for various J-related groups.


It is unfortunate that there is no progress in the phylogeographic assessment of R1a in this paper. There have been substantial discoveries of SNPs within this haplogroup as a result of commercial testing; however there is clearly an ascertainment bias in the newer discoveries, as almost all these SNPs have been detected in Europeans. The new paper confirms the high levels of Y-STR variance in India, Pakistan, and Iran. Together with the cornucopia of related paragroups in Iran, there is little doubt that this haplogroup originated in the general area of Central/South Asia.


Personally, as I have stated before, I would relate this R1a with Neolithic peoples living east of the Caspian, in contrast to the R1b bearers who lived west and south of it. These two populations came under the influence of the Indo-Europeans and spread in different directions. The Indo-Iranians were then initially the mixed descendants of the Indo-Europeans and the R1a old agricultural population, and were formed in the territory of the Bactria-Margiana Archaeological Complex. 


This also explains the contrast between Iranian and Armenian groups: the latter mostly lack the R1a lineage, contrasting with all Iranian groups (even their Kurdish neighbors) who possess it. Conversely, Iranian groups, and especially eastern Iranians and Indo-Ayrans lack the R1b lineage. This is due to the fact that neither R1a nor R1b were originally part of the Indo-European community, but their geographical position was such that they came under the influence of the Indo-Europeans when the latter began their expansion.


UPDATE VI: I have created my own dendrogram using the Y-haplogroup frequencies and the hclust package of R (default parameters):

From top to bottom, one can identify some clusters:

• Eastern Europe, further broken down into Balkans and Slavic+Hungary
• West Asian/Caucasus
• Iranian Proper
• Arab

These correspond largely to the clusters identified by the authors, with India and the Turkmen sample emerging as the clear outliers. I omitted the Ethiopian samples, since E-M78 was not resolved phylogenetically, causing the Ethiopians to group with the likely E-V13 from the Balkans.

UPDATE VII: I have also run MCLUST over the haplogroup frequency data over the MDS representation of the distance matrix. The maximum number of 10 clusters occurred with 5 MDS dimensions retained. Population assignments in the 10 clusters can be found in the table below:

Iran/Azerbaijan_Gharbi+Tehran_(Assyrian)	1
Iran/Lorestan_(Lur)	1
Iran/Tehran_(Armenian)	1
Iran/Azerbaijan_Gharbi_(Azeri)	2
Iran/Hormozgan_(Bandari+Afro-Iranian)	2
Iran/Hormozgan/Qeshmi	2
Iran/Khorasan_(Persian)	2
Iran/Kurdistan_(Kurd)	2
Iran/Sistan_Baluchestan_(Baluch)	2
Pakistan	2
Iran/Fars+Isfahan_(Persian)	3
Iran/Gilan_(Gilak)	3
Iran/Yazd+Tehran_(Zoroastrian)	3
Turkey/Central	3
Turkey/East	3
Turkey/West_	3
Iran/Golestan_(Turkmen)	4
India	4
Iran/Khuzestan_(Arab)	5
Egypt_(Arab)	5
Iraq/Baghdad	5
Oman	5
Saudi_Arabia	5
Tunisia	5
United_Arab_Emirates	5
Iran/Mazandaran_(Mazandarani)	6
Iran/Yazd_(Persian)	6
Balkarian	6
Georgia	6
Albania	7
Greece	7
Bosnia	8
Croatia	8
Slovenia	8
Czech_Republic	9
Hungary	9
Poland	9
Ukraine	9
Iraq_(Marsh_Arab)	10
Qatar	10
Yemen	10

We can ignore cluster #4 which consists of the two outliers (India + Turkmen). The rest of the clusters seem relatively coherent. Notice, for example, the Arabian cluster #10, Balkan cluster #8, Eastern European cluster #9, Greek-Albanian cluster #7, Mixed Arab cluster #5.

PLoS ONE 7(7): e41252. doi:10.1371/journal.pone.0041252

Ancient Migratory Events in the Middle East: New Clues from the Y-Chromosome Variation of Modern Iranians

Viola Grugni et al.


Knowledge of high resolution Y-chromosome haplogroup diversification within Iran provides important geographic context regarding the spread and compartmentalization of male lineages in the Middle East and southwestern Asia. At present, the Iranian population is characterized by an extraordinary mix of different ethnic groups speaking a variety of Indo-Iranian, Semitic and Turkic languages. Despite these features, only few studies have investigated the multiethnic components of the Iranian gene pool. In this survey 938 Iranian male DNAs belonging to 15 ethnic groups from 14 Iranian provinces were analyzed for 84 Y-chromosome biallelic markers and 10 STRs. The results show an autochthonous but non-homogeneous ancient background mainly composed by J2a sub-clades with different external contributions. The phylogeography of the main haplogroups allowed identifying post-glacial and Neolithic expansions toward western Eurasia but also recent movements towards the Iranian region from western Eurasia (R1b-L23), Central Asia (Q-M25), Asia Minor (J2a-M92) and southern Mesopotamia (J1-Page08). In spite of the presence of important geographic barriers (Zagros and Alborz mountain ranges, and the Dasht-e Kavir and Dash-e Lut deserts) which may have limited gene flow, AMOVA analysis revealed that language, in addition to geography, has played an important role in shaping the nowadays Iranian gene pool. Overall, this study provides a portrait of the Y-chromosomal variation in Iran, useful for depicting a more comprehensive history of the peoples of this area as well as for reconstructing ancient migration routes. In addition, our results evidence the important role of the Iranian plateau as source and recipient of gene flow between culturally and genetically distinct populations.

Link

First look at Turkish and Kyrgyz data from Hodoğlugil & Mahley (2012)

The authors of the recent paper on Turkish population structure were kind enough to share their data with me. I will be sure to use this data in future experiments, such as the ChromoPainter and fastIBD analysis of Balkans/West Asia, as well as a ChromoPainter analysis of Altaic speakers, following on the footsteps of my recent analysis of Afroasiatic speakers.

PCA


As a first step, after processing the new data, I carried out a PCA analysis (in smartpca with no outlier removal iterations), combined with various Turkic groups, as well as a few neighbors of Anatolian Turks, combining data from the literature and the Dodecad Project.

The Turkic cline from East to West Eurasia, observed by myself and others in various experiments is again evident.

The blowup of the above, focusing on the West Eurasian portion (top right) is easier to read:

As always, population labels are placed in the average position of each population. So, for example, the Behar et al. Iranians_19 sample is shifted to the left, because of the existence of a few African admixed individuals in this sample. The Iranian_D sample of Project participants seem to lack this admixture.

Also, note that since there is no South Asian reference in this first experiment, Iranians overlap with Turks along the first two dimensions. As we've seen in the Dodecad Project, both Iranians and Anatolian Turks are "eastward-shifted" relative to other West Eurasians, but the former have a strong South Asian- and the latter a Central Asian- tendency.

The new Kyrgyz sample falls between the Kazakh and the Altai along the cline, and is more "eastern" compared to the Uygurs and Uzbeks, and more "western" compared to Altai, Tuva, and Dolgans.

Kayseri and Istanbul Turks overlap with Behar et al. Turks as well as the Turkish_D sample. The Aydin sample appears to be more heterogenous, with a more eastern overall center of weight. More on this below.

ADMIXTURE


I also carried out a K=3 ADMIXTURE analysis of the dataset.

Below are the population portraits for the three new Turkish samples, as well as the Kyrgyz sample:

It is obvious that many Turks have low levels of Asian admixture, lacking in their geographical neighbors, but this is quite variable on an individual basis.

UPDATE (17 Feb):


I have also assessed the new data with the K12b calculator. Below are the normalized median proportions.

Variation in four Central Anatolian settlements

I had linked to the conference versions of this work in ISABS 2007 and AAPA 2008. Now there is a new paper in American Anthropologist on the topic of variation in four Central Anatolian settlements with very different origins. As I've mentioned before, this is a great illustration of the problem of uncritically treating modern Anatolian Muslim samples as representatives of the Neolithic population, for at least two reasons:

1. Modern Anatolian Muslims are only a part of the recent population of Anatolia, with a great part of the Christian population exchanged, killed, or forced to leave.
2. Modern Anatolian Muslims are ethnically, linguistically and religiously heterogeneous, and many of them have historical memories of descent from elsewhere, e.g., the Balkans, the Caucasus, or Central Asia

So, I have always been skeptical of grand reconstructions of prehistory based on treating modern Anatolian Muslims as little modified descendants of the Neolithic population. I don't doubt that a great deal of their ancestry, averaged over broad scales, is Neolithic Anatolian, but extracting this ancestry from the mosaic of heterogeneous Muslim groups inhabiting the modern Republic of Turkey is no trivial task.

From the current paper, in support of point #1:

Even though the data from official Ottoman records and other sources such as local church accounts are contradictory (Charanis 1972; Vryonis 1986), it is clear that Greek and Armenian peoples comprised the majority, or at least a very sizeable minority, of the late-19th-century Anatolian population (Finkel 2005; Levy 2002; Shaw et al. 1976). Muslim Turkic and Kurdish groups from different ancestral clans and, more importantly, from different sects of Islam made up the largest remaining part of the population (Cahen 1968, 2001; Finkel 2005). Therefore, the population of the Ottoman Empire came from different ancestral backgrounds, lived together, and constructed communities based on their religious affiliations.

In support of point #2, the authors have studied the inhabitants of a Central Anatolian region:

We worked in four geographically proximate Central Anatolian settlements located southeast of Ankara (see Figure 1) to test the abovementioned hypotheses and elucidate the regional complexity of Anatolian population history. Because of the current political sensitivities concerning ethnic–religious identity in Turkey, especially those relating to the Alevis and Kurds, the names of the specific settlements we visited are not identified in this article. Instead, for the sake of clarity, pseudonyms are used. This region, which we refer to as “Yuksekyer,” was selected for the study because, based on oral traditions and available historical records, it is home to linguistically, ethnically, and religiously distinct groups that live in close geographic proximity to one another. To assess the possible regional, religious, and ethnic differentiation in Central Anatolia, we collected additional samples from a settlement from Kizilyer, another region located about 500 kilometers east of Yuksekyer, the inhabitants of which are predominantly Alevi Turks.

"Yuksekyer" is 2,500 sq. km in size, so it's quite small in the broader Anatolian context. Here are details on the studied settlements:

Merkez is the current political and bureaucratic center with about 6,500 inhabitants, making it the most populous settlement in the region (Devlet İstatistik Enstitüsü 2001). It was probably founded by Cerkez (Circassian) people who had migrated there from the Caucasus region in the 14th century.

...

The inhabitants of the oldest known settlement, Eskikoy, claim a pre-Ottoman Karaman ancestry that traces back to a Turkic population that occupied the Konya region during the 13th century (Finkel 2005). To date, no historical records that confirm the connection between Eskikoy and its putative Karaman origin have been found. However, Ottoman records mention the presence of the Eskikoy settlement in the Yuksekyer region and place its foundation at around 1500 C.E.

...

The residents of Gocmenkoy identify themselves with the Afsar clan of the Oguz tribe, to which the Kayi and Turkmen lineages also belong (Cahen 1968). Their oral history, supported by local historical records, indicates that these people came from Central Asia in the 16th century.

...

The Kurdish-speaking inhabitants of Dogukoy were the last immigrants to populate Yuksekyer. They purportedly came into the region around 200 years ago from southeastern Turkey.

...

For comparative purposes, we also collected 30 additional samples from another Central Anatolian region, Kizilyer, which lies about 500 kilometers east of Yuksekyer. This region is roughly comparable to Yuksekyer region in its size and population density.

There are a lot of interesting genetic data, but I will focus on the Y-haplogroup profiles of the different populations:

A few things stand out:

• Haplogroup L is limited to Gocmenkoy. L is divided into informative subclades and is one of the less studied and more mysterious Caucasoid haplogroups; the authors erroneously state that it is more frequent in East Eurasians. In the Gocmenkoy sample it could very well be the legacy of Turkicized Central Asian Iranian speakers in which it is found at a high frequency. Most of the haplogroup Q is also found in Gocmenkoy and this may represent a genuine Turkic element.
• Haplogroup J1 is largely limited to Merkez; this is not surprising as Merkez is said to have been founded by Circassians, and J1 occurs at a substantial frequency in parts of the Caucasus
• Haplogroup N is mostly found in Eskikoy, and this is also a likely marker of Central Asian Turkic groups
• I find the high frequency of J2a (64.5%) in Dogukoy, the Kurdish settlement to be noteworthy. This haplogroup is also probably found at a high frequency in Parsis (although technically only J was studied in the relevant study), and I've noted before that a high J2/J1 ratio contrasts West Asian Indo-Europeans from Semitic groups. J2a also occurs at a high frequency in Indian upper caste populations, whereas it is virtually absent in low castes and tribals.

The fact that the names of the settlements had to be obfuscated by the authors speaks volumes about the political sensitivity of the subject. Hopefully this line of research will be allowed to continue.

Related:

• How Turkish are Anatolians? revisiting the question

American Anthropologist
Volume 113, Issue 1, pages 116–131, DOI: 10.1111/j.1548-1433.2010.01310.x

Biological Ancestries, Kinship Connections, and Projected Identities in Four Central Anatolian Settlements: Insights from Culturally Contextualized Genetic Anthropology

Ömer Gokcumen et al.

ABSTRACT  Previous population genetics studies in Turkey failed to delineate recent historical and social factors that shaped Anatolian cultural and genetic diversity at the local level. To address this shortcoming, we conducted focused ethnohistorical fieldwork and screened biological samples collected from the Yuksekyer region for mitochondrial, Y chromosome, and autosomal markers and then analyzed the data within an ethnohistorical context. Our results revealed that, at the village level, paternal genetic diversity is structured among settlements, whereas maternal genetic diversity is distributed more homogenously, reflecting the strong patrilineal cultural traditions that transcend larger ethnic and religious structures. Local ancestries and origin myths, rather than ethnic or religious affiliations, delineate the social boundaries and projected identities among the villages. Therefore, we conclude that broad, ethnicity-based sampling is inadequate to capture the genetic signatures of recent social and historical dynamics, which have had a profound influence on contemporary genetic and cultural regional diversity.

Link

Some ADMIXURE estimates on Eurafricans

Here is an ADMIXTURE run on the Xing et al. (2010) dataset using markers with less than 1% genotyping no-calls. Populations (left-to-right) are: Alur, Hema, Luhya (from Africa), and CEU, Tuscan, Slovenian, Urkarah, and Kurdish (from Western Eurasia).

Most West Eurasians have no trace of Sub-Saharan ancestry, except Kurds (estimated here at 2.3% overall, range: 1-3.6%), and a couple of Slovenians (4.5% and 1.6%, overall 0.2%). The fairly uniform distribution in Kurds suggests to me that this is an ancient phenomenon reflecting gene flow between the Near East and Africa and not a recent phenomenon.

The Hema are an African population exhibiting West Eurasian affinities (11.0%, range: 4.2-15.2%). Consult my comments on the original Xing et al. article on some information on these Nilo-Saharan pastoralists from the Democratic Republic of Congo.