The formation of the Middle Eastern Cline

[IronSide]

In the PCA of West Eurasians, there are three clines of ancestry, one is mainland Europe another is the Middle East, and a third between them for Meditteranean and Jewish populations.





The European cline formed as a mixture of Steppe BA populations and European Late Neolithic/Chalcolithic populations, did the Middle Eastern cline form similarly as a mixture of two populations to varying degrees? if so then what are they?

The first fact is that it is a north-south cline, the northernmost populations are north Caucasians (Adygei, Balkar, Chechen, Georgian Ossetian, Lezgin, Kumyk, Abkhasian), while the southernmost seem to be Bedouins and Arabians.

If we take Europeans as an analogy, the southernmost and northernmost populations are a proxy for the two founder populations, is it the same in the Middle East?

Ancient DNA has already documented a mixture between Iran Chalcolithic on the one hand and Levant Neolithic/Anatolian Neolithic populations on the other, resulting in the Levant BA and Anatolia BA, but they're not the same as modern Turks and Levantines, and north Caucasians have extra Steppe ancestry compared to EBA populations, so this large Early BA migration from Iran/Caucasus did contribute but not fully forming the cline between north and south.

My hypothesis is that The Middle East cline formed through three waves of migrations:

1- Anatolian and Levantine farmers migrate to the Caucasus and Iran, combining with CHG to form Armenian and Iranian BA populations.

2- these populations then back-migrate to Anatolia and the Levant, forming the Levantine and Anatolian Early BA populations.

3- in the Middle Bronze Age, Armenian Middle/Late BA forms with additional EHG ancestry, hinting that Steppe ancestry arrived at the region, North Caucasians are the closest proxy population, this group then moved everywhere, probably like this:

Anatolia: Anatolia_BA + Armenia_MLBA
Iran: Iran_Chalcholithic + Armenia_MLBA
Levant: Levant_BA + Armenia_MLBA

for Arabia, my hypothesis is a Levant Iron Age migration combining with the ancestral southern population, which is probably Levantine Neolithic but shifted towards the Natufians, I'll call it Arabian Neolithic.

Arabia: Levant_IA + Arabian_Neolithic

Armenia has shifted south again, probably after the Assyrian campaigns to Urartu:

Armenia: Armenia_MLBA + Levant_IA

I'm comfortable with this scheme because it seems to agree with the modern populations positions on the PCA compared to their previous state, they're all shifted to north Caucasians.

The biggest hole in all of this is Mesopotamia, but it should be intermediate between the Levant and Iran, this northern migration may have reached Mesopotamia and had a role in forming the new Kassite Babylonian empire.

I have some theories for linguistic association for all of these movements but in another post.

[Govan]

Mesopotamia and Eastern Anatolia are the most mixed ME regions , i think even back then they were big mishmashed points of mixing.

[Boreas]

Mesopotamia and Eastern Anatolia are the most mixed ME regions , i think even back then they were big mishmashed points of mixing.

I think that Western Anatolians are more mixed then Easterns.

[Ygorcs]

I think that Western Anatolians are more mixed then Easterns.

I'd expect Western Anatolia to have been less mixed than Eastern Anatolia (read: less subject to foreign invasions of highly genenetically distinct populations from without the Mediterranean Basin), but it's probable that this changed after the Turkic immigration and the Ottoman Empire. Turkic immigration seems to have impacted Western Anatolia much more heavily, and later the significant flows from the Balkans and other regions of the transcontinental empire seem to have concentrated more on the western, Mediterranean Anatolia.

[Ygorcs]

IronSide, I think your model is probably a bit simpler than the complex reality of the multiple movements of populations and admixture events in the Early Neolithic-to-Early Iron Age Middle East, but IMHO you got the "essence" of the main genetic/demographic trends very well.

[Boreas]

I'd expect Western Anatolia to have been less mixed than Eastern Anatolia (read: less subject to foreign invasions of highly genenetically distinct populations from without the Mediterranean Basin), but it's probable that this changed after the Turkic immigration and the Ottoman Empire. Turkic immigration seems to have impacted Western Anatolia much more heavily, and later the significant flows from the Balkans and other regions of the transcontinental empire seem to have concentrated more on the western, Mediterranean Anatolia.

East Anatolia is part of main base of South Caucausian-Anatolian gene pool, so their are more pure. Also as you noticed, Turkic admixture is higher in West. This brings South Asian and Siberian admixtures and with Balkan migration and Turkification of Local West part has also significant European admixture.


@IronSide

Bedouin Structure seems extremly wide. If you have knowledge, please share.

[raspberry]

An upcoming paper has found Iranian and Causian ancestry in North Mesopotamia dating back to the early Bronze Age.

Can you give a link to the abstract or something? Thank you.

[raspberry]

http://www.icaane2018.vorderas-archa....de/index.html
Central to this question is the archaeological site of Arslantepe, strategically located at the Upper Euphrates, the nexus of all three regions. Arslantepe also developed one of the first state societies of Anatolia along with advanced metal-technologies. Archaeological research suggests that conflicts with surrounding groups of pastoralists affiliated to the Caucasus might have contributed to the collapse of its palatial system at the end of the Chalcolithic period (4th millennium BC). To test if these developments were accompanied by genetic changes, we generated genome-wide data from 18 ancient individuals spanning from the Late Chalcolithic period to the Early Bronze Age of Arslantepe. Our results show no evidence for a major genetic shift between the two time periods. However, we observe that individuals from Arslantepe are very heterogeneous and differentiated from other ancient western and central Anatolians in that they have more Iran/Caucasus related ancestry. Our data also show evidence for an ongoing but also recent confluence of Anatolian/Levantine and Caucasus/Iranian ancestries, highlighting the complexity of the Chalcolithic and Bronze Age periods in this region.
In case you didn't know this, Arslantepe is accepted by relevant experts as an extension of North Mesopotamia.

Great. Arslantepe was once a Hittite city and the abstract sounds promising. Soon those people who called the previous Hittite samples Hattians will now be punched in their ignorant faces once more by another paper.

[raspberry]

The site predates the Hittites.

That's even better.

[IronSide]

This is a wonderful online tool made by poi from Anthrogenica. Credits:

This tool would not be possible without:
nMonte Logic: @huibregts
Global25 Datasheet: @Davidski

further information: https://anthrogenica.com/showthread....y-from-the-web

I modeled some populations from the Near East, my procedure is systematic, I take a population and try to model it with the most recent sample available, it usually gives a poor fit, then I try to add "basic" populations (EHG, CHG, Onge, Natufian .. etc) and see if the error dropped and that it requires some ancestry from that "basic" type (this is an awkward approximation for the behaviour of a D-stat), I go through all types, then I try to model it with a logical proximal source that contains these basic types.

Saudi: 82% Levant_BA + 3% Dinka + 15% Natufian
fit=4.3
this is not a very good fit, (Natufian + Dinka) is approximate to whatever population lived in Arabia before the huge Levant_BA colonization which likely brought Semitic languages. "BedouinB", Yemenite Jews and many highland Yemeni groups are modeled similarly to Saudi.

Somali: 56% Dinka + 5% Iran_N + 39% Natufian
fit=5.2
this is also a terrible fit, but the best I could reach, its likely because Natufian is not an exact replica of the original population of Arabia, the ghost population that combines with Levant_BA in Saudis and Yemenis, this IMO says that Arabia was populated by Cushitic speakers before Semitic peoples arrived, Iran_N drops the fit so I added it, it must then have existed in Arabia with this "Natufian" like thing.

Ethiopian: 20% Levant_BA + 80% Somali
fit=2.2
a beautiful result, Ethiopians are Semitic speakers, and so are expected to have more Levant_BA ancestry, from southern Arabia perhaps, as most of it's ancestry is Levant_BA like, the substrate population must have been Cushitic speakers like Somalis.

BedouinA: 26% Armenia_MLBA + 7% Dinka + 63% Levant_BA + 4% Natufian
fit=2.3
BedouinA is a label for a group in the Negev, who migrated from southwestern Saudi Arabia, many tribes including mine are modeled similarly to this group, the obvious difference is Armenia_MLBA ancestry, which speaks for a northern origin of these tribes who have a historical origin in the Kingdom of Saba.

Palestinian: 30% Armenia_MLBA + 5% Dinka + 65% Levant_BA
fit=2.8
Armenia_MLBA ancestry is a legacy of the Hurrians from the Amarna age, when the New Kingdom of Egypt dominated Canaan, Levant_BA are the Canaanites, some Palestinian samples (maybe from coastal areas?) are modeled better with Mycenaeans??, maybe Philistines and Danites, but the signal can also be modeled with Anatolia_BA. Jordanians are modeled similarly to Palestinians.

Lebanese: 39% Anatolia_BA + 17% Armenia_MLBA + 44% Levant_BA
fit=2.5
Sidon_BA had more Anatolian? and CHG ancestry than Jordan_EBA so not all this Anatolia_BA is "Hittite" but some of it may truly be Anatolian, the Sidon Canaanite is not available, but we observe the same pattern of an Armenian Levantine Anatolian mixture, I used Lebanese Christians in this model, Muslims are very similar, with 1% African but also varying degrees of North Caucasus ancestries.

Assyrian: 22% Anatolia_BA + 30% Armenia_MLBA + 16% Levant_BA + 33% Iran_ChL
fit=2.5
There are no Syrian or Iraqi samples, but Assyrians may capture the same admixture history. The model looks cool, don't know if its real, I feel it could be simplified.

Armenian: 34% Anatolia_BA + 47% Armenia_MLBA + 5% Levant_BA + 14% Iran_ChL
fit=2.1
The northern Middle East has a really complex history of migrations, most of it may have been driven by the Late Bronze Age collapse, Armenians differ from the Late Bronze Age is having more Western ancestry, and modern Anatolian Turks have more Eastern ancestry.

Eastern Turkey (Adana & Keysari): 41% Anatolia_BA + 45% Armenia_MLBA + 14% Uygur
fit=3
You could argue its just Armenian plus Central Asian Turk, Turkmen instead of Uygur also works.

Western Turkey (Aydin & Balikesir): 36% Anatolia_BA + 34% Armenia_MLBA + 30% Uygur
fit=3.25
Turkish migration was stronger to Western Turkey, the main pattern is strong bidirectional collapse of ancestry in Anatolia and Armenia, after the distruction of the Hittite Empire, there is evidence for an Eastern type of pottery in places like Gordion, the capital of the Phrygian state.

Iranian Persian: 27.5% Armenia_MLBA + 4% Levant_BA + 27% Pashtun + 42% Iran_ChL
fit=2
The theoritical homeland of Proto Indo-Iranian is the Sintashta culture, both Pashtun and Armenia_MLBA have significant amounts from Steppe_MLBA, I added Pashtun because modern Iranians have ASI ancestry that wasn't present in the Chalcholithic, so I thought a proximate source could also correspond to a group that carried Iranian languegs, but the linguistic affinities for Armenia at this time are also unclear.

Kurdish: 50% Assyrian + 50% Iranian_Persian
fit=1.6
What a clean result.

What else ? Georgians, North Caucasians, and Azeris, later maybe.

[davef]

Thanks, keep going with this when you have the time! Lebanon is pretty Anatolian.

[brick]

This is a wonderful online tool made by poi from Anthrogenica. Credits:

This tool would not be possible without:
nMonte Logic: @huibregts
Global25 Datasheet: @Davidski

Good work but...

nMonte is a good tool but Huijbregts always repeats that results under 10% must be interpreted with great caution because may not be accurate.

Global25 Datasheet is less good because is the product of the various agendas of its creator and none can check if Global25 Datasheet's averages are accurate. Davidski is accurate only with the ethnicities that interest him.

[Wigster]

Interesting modelling but I have some other models in mind regarding West Asians.

"distance%=4.159"

Saudi

Levant_BA,68.2
Natufian,23.4
Seh_Gabi_ChL,6.4
Ethiopia_4500BP,1.2
Dinka,0.4
Yoruba,0.4

Arabians seem mostly like natives who got Semitized. The Arabian Penninsula was probably something like Jordan_BA + significantly more Basal. There's more Iran_Chl in addition to the Jordan_BA sample in modern day Saudis though, possibly Mesopotamian or from later Levantine admix. But the fit is rather bad, probably because it has too much Anatolian_N/WHG-related.

"distance%=1.4926"

Lebanese_Christian

Sidon_BA,77.8
Seh_Gabi_ChL,10.2
Anatolia_BA,7.6
Hajji_Firuz_BA,4.4

Lebanese Christians seem pretty straightforward. They owe vast majority of their ancestry to the Bronze Age Canaanites with smaller addition of Zagros_Chl like admix. Hajji Firuz_BA is the intrusive 50% Yamnaya sample from NW Iran. This kind of admixture probably came with the Hurrians and other Mesopotamians and not from the Iranian plateau.
Iron Age Levantines such as the Phoenicians and Arameans probably had this extra Armenia_EBA/Iran_Chl like admix already.

"distance%=1.1723"

Palestinian

Sidon_BA,45.6
Levant_BA,30.4
Hajji_Firuz_BA,10.4
Seh_Gabi_ChL,6.6
Ethiopia_4500BP,4.6
Yoruba,2.4

Palestinians are significantly more shifted to the south in comparison with Lebanese Christians. Probably due to Arab admix. Also higher Steppe, probably a relic of the Crusades.

"distance%=1.0343"

Assyrian

Sidon_BA,43.8
Seh_Gabi_ChL,27
Armenia_EBA,16.4
Hajji_Firuz_BA,9.6
Anatolia_BA,3.2

Good fit for Assyrians, but the samples are not particularly relevant though ofc. But this is the more logical fit without access to Mesopotamian and Akkadian aDNA. Somewhat similar to the Lebanese Christians but in different numbers. Assyrians probably owe 50%+ of their heritage to Sumerians, Hurrians and other Mesopotamians who were probably somewhat intermediate between Sidon_BA and Iran_Chl.

"distance%=1.2939"

Armenian

Armenia_EBA,35.4
Hajji_Firuz_ChL,28.2
Sidon_BA,24
Anatolia_BA,9.2
Hajji_Firuz_BA,3.2

Interestingly Armenians score very little Steppe. They probably got Indo-Europeanized by people who had very little Steppe input themself probably on the same level as the Mycenaean samples.

"distance%=1.0421"

Kurdish

Seh_Gabi_ChL,39.2
Sidon_BA,29.4
Turkmenistan_IA,24.6
Anatolia_BA,4.2
Hun_Tian_Shan,2.6

Kurds are considerably Central Asian shifted. The R1a-Z93 Iron Age sample from Turkmenistan is 50% Steppe with the rest being BMAC-like. Another interesting thing is the high Levantine input just like with Armenians. Some of it might be ancient Mesopotamian and Elamite. Not sure how much Kurds would've intermixed with Iraqi Arabs throughout the centuries but it might act as a proxy for Arab + Anatolian admix as that kind of mix would resemble the Sidon samples.

Iraqi and Syrian Arabs are not represented in the spreadsheet, but I presume Syrians to be close to Lebanese but shifted towards both Assyrians and Saudis.
Iraqis are basically Saudi/Iranian mixes.

[IronSide]

Interesting modelling but I have some other models in mind regarding West Asians.

"distance%=4.159"

Saudi

Levant_BA,68.2
Natufian,23.4
Seh_Gabi_ChL,6.4
Ethiopia_4500BP,1.2
Dinka,0.4
Yoruba,0.4

Arabians seem mostly like natives who got Semitized. The Arabian Penninsula was probably something like Jordan_BA + significantly more Basal. There's more Iran_Chl in addition to the Jordan_BA sample in modern day Saudis though, possibly Mesopotamian or from later Levantine admix. But the fit is rather bad, probably because it has too much Anatolian_N/WHG-related.

"distance%=1.4926"

Lebanese_Christian

Sidon_BA,77.8
Seh_Gabi_ChL,10.2
Anatolia_BA,7.6
Hajji_Firuz_BA,4.4

Lebanese Christians seem pretty straightforward. They owe vast majority of their ancestry to the Bronze Age Canaanites with smaller addition of Zagros_Chl like admix. Hajji Firuz_BA is the intrusive 50% Yamnaya sample from NW Iran. This kind of admixture probably came with the Hurrians and other Mesopotamians and not from the Iranian plateau.
Iron Age Levantines such as the Phoenicians and Arameans probably had this extra Armenia_EBA/Iran_Chl like admix already.

"distance%=1.1723"

Palestinian

Sidon_BA,45.6
Levant_BA,30.4
Hajji_Firuz_BA,10.4
Seh_Gabi_ChL,6.6
Ethiopia_4500BP,4.6
Yoruba,2.4

Palestinians are significantly more shifted to the south in comparison with Lebanese Christians. Probably due to Arab admix. Also higher Steppe, probably a relic of the Crusades.

"distance%=1.0343"

Assyrian

Sidon_BA,43.8
Seh_Gabi_ChL,27
Armenia_EBA,16.4
Hajji_Firuz_BA,9.6
Anatolia_BA,3.2

Good fit for Assyrians, but the samples are not particularly relevant though ofc. But this is the more logical fit without access to Mesopotamian and Akkadian aDNA. Somewhat similar to the Lebanese Christians but in different numbers. Assyrians probably owe 50%+ of their heritage to Sumerians, Hurrians and other Mesopotamians who were probably somewhat intermediate between Sidon_BA and Iran_Chl.

"distance%=1.2939"

Armenian

Armenia_EBA,35.4
Hajji_Firuz_ChL,28.2
Sidon_BA,24
Anatolia_BA,9.2
Hajji_Firuz_BA,3.2

Interestingly Armenians score very little Steppe. They probably got Indo-Europeanized by people who had very little Steppe input themself probably on the same level as the Mycenaean samples.

"distance%=1.0421"

Kurdish

Seh_Gabi_ChL,39.2
Sidon_BA,29.4
Turkmenistan_IA,24.6
Anatolia_BA,4.2
Hun_Tian_Shan,2.6

Kurds are considerably Central Asian shifted. The R1a-Z93 Iron Age sample from Turkmenistan is 50% Steppe with the rest being BMAC-like. Another interesting thing is the high Levantine input just like with Armenians. Some of it might be ancient Mesopotamian and Elamite. Not sure how much Kurds would've intermixed with Iraqi Arabs throughout the centuries but it might act as a proxy for Arab + Anatolian admix as that kind of mix would resemble the Sidon samples.

Iraqi and Syrian Arabs are not represented in the spreadsheet, but I presume Syrians to be close to Lebanese but shifted towards both Assyrians and Saudis.
Iraqis are basically Saudi/Iranian mixes.

Good models, but regarding Palestinians:

Canaanites were the indigenous population of the Southern Levant during the 2nd millennium BCE. Their genetic origin and impact on modern populations have recently started to unravel following an analysis of a Lebanese sample. To study Canaanites from other Levant regions and their genomic heritage in the broader Middle East, we sequenced five petrous bones from Megiddo, Israel, dated to the Middle/Late Bronze Age (BA) transition (≈3.5 KYA). We enriched the DNA for approximately 1.2 million SNP targets, followed by sequencing at coverage >0.25x. Using a combination of statistical tools (PCA, f-statistics, ADMIXTURE, qpAdm), we found that the Megiddo samples can be modeled as a mixture of earlier samples from the Levant and Iran, the latter possibly representing migration via Armenia. The Megiddo samples showed high similarity to older Levant BA samples, as well as to a later Iron Age (IA, ≈3 KYA) sample that we sequenced from Abel Beth Maacah in Northern Israel. The genomes of modern native Levantine populations trace ≈60% of their ancestry to IA Canaanites, ≈10% to Eastern Africa, and the remaining to less well characterized sources, possibly related to Iran. The genomes of Ashkenazi Jews can be modeled as ≈55% BA Canaanites and ≈45% Neolithic Central Europeans, and those of Iraqi Jews as ≈70% BA Canaanites and ≈30% Neolithic Iranians. To validate the results, we developed a novel extension of ChromoPainter that can take advantage of the information in linked SNPs to paint ancient chromosomes and model their ancestry. Our results confirm previous findings regarding the mixed Levantine-Iranian ancestry of BA Canaanites, and suggest remarkable continuity in the region throughout the Bronze and Iron Ages. Using existing and new methods, we characterized the ancestry of modern Middle-Eastern populations as a combination of pre-existing groups from the Middle-East and beyond."

Palestinians and Jordanians should have up to 30% ancestry from a probable Armenian source that didn't exist in the Jordan_EBA samples, there seems to have been a north-south cline from Sidon to Jordan of decreasing Iran_ChL ancestry, but you can see that a northern shift occurred in both locations, historically at this time Hurrians settled Canaan in large numbers, in your modelling, Palestinians become more like Sidon_BA, should this be interpreted as a movement from the Lebanese coast ? or an artifact of caused by a probable Caucasus source ?

I'm skeptical of any model of Armenians that doesn't incorporate Armenia_MLBA, what, did they just vanish ?

[Wigster]

Good models, but regarding Palestinians:



Palestinians and Jordanians should have up to 30% ancestry from a probable Armenian source that didn't exist in the Jordan_EBA samples, there seems to have been a north-south cline from Sidon to Jordan of decreasing Iran_ChL ancestry, but you can see that a northern shift occurred in both locations, historically at this time Hurrians settled Canaan in large numbers, in your modelling, Palestinians become more like Sidon_BA, should this be interpreted as a movement from the Lebanese coast ? or an artifact of caused by a probable Caucasus source ?

I'm skeptical of any model of Armenians that doesn't incorporate Armenia_MLBA, what, did they just vanish ?

"Modern day Levantines" could mean many people. But it's clear Palestinians have more "Red Sea" like input in comparison with other Levantines. I think it's for sure related to the Islamic expansion because the Jordan_BA samples seem like a "dead end". Most Semitic speakers with the exception of Saudis/Bedouins and Ethiopians prefer the Sidon samples when modelled. It seems like they were either mixed with people who were more Basal, probably from the Arabian penninsula or there was a complete population replacement with Sidon-like people. Even modern day Jordanians prefer the Sidon samples over the ones from Bronze Age Jordan. I think it's safe to presume some or most of the Jordan_BA we see in Palestinians and Jordanians to have come with Arabs.

As an example removing Levant_BA and adding Bedouins for Palestinians I think we get a more accurate composition

"distance%=1.133"

Palestinian

Sidon_BA,61.8
BedouinB,16.4
Hajji_Firuz_BA,7.6
Seh_Gabi_ChL,7.6
Ethiopia_4500BP,4.6
Yoruba,2

Basically a Canaanite core with extra Iran_Chl input from the Mesopotamian expansions, extra Euro from the Crusades(?) or possibly earlier Indo-Europeans, some SSA from the Slave trade and a decent amount of Arabian probably from the Islamic conquests and onward.

Btw I decided not to use Armenia_MLBA because it's a very low coverage sample.

[brick]

Sidon_BA has been removed from Global25 Datasheet. If it has been removed, there must be a reason.

[Angela]

Well, I would hope someone takes a look at it to see that it just isn't that the models don't turn out the way the author wants them to if Sidon Bronze Age are included.

People have to stop being so gullible.

[markod]

Doing my own experiments I had noticed a while ago that fits for Levantine populations can be significantly improved by adding Egyptians as an outgroup. I believe the reason for this is that a population presumably associated with the Iberomaurasians brought Afroasiatic languages to the Middle East. Other North Africans do not improve the fit to the same extnt, while other Levantines makes it signifcantly worse, so I suspect there might be a component modal in present day Egyptians which arrived in the osuthern Levant. Take for example this model for Palestinians, which I based on yours and which is by no means d meant to be taken as definitive. MLBA here represents the highland West Asians who spread with the metallurgical innovations, Myceneans the Philistines and other Indo-Europeans, Levant_BA the settled population of the Levant and Egyptians the presumable Proto-Semites who migrated across the Sinai:

Palestinian: 25% Armenia_MLBA + 33% Egyptian + 34% Levant_BA + 8% Mycenean
Fit = 1.8

A similar model for Bedouin_A which doesn't like the Mycenean input:

Bedouin_A: Armenia_MLBA 15% + Egyptian 53% + Levant_BA 30%
Fit = 1.5

This also eliminates residual African admixtures usually attributed to slave trade and such.