The Archaeogenetics of the Middle East

[Blaze86420]

I'd like to revive this thread, if you don't mind Sumskilz. I was going over this study that you linked earlier: http://journals.plos.org/plosgenetic...l.pgen.1003316

I'm struggling to understand this graph here:

Spoiler Alert, click show to read: 

So if I understand this study correctly, it's saying that minorities in the Levante cluster close to Jews, Cyptriots, and Europeans while Jordanians and Palestinians cluster close to Saudis. Wouldn't you expect Palestinians and Jordanians to cluster right between Saudis and Levantine minorities? Unless I misunderstood you, you said that Palestinians typically have 50% Levantine ancestry, and the rest being primarily Arabian. And I have absolutely no clue what the tree chart is saying.

[norse]

I'd like to revive this thread, if you don't mind Sumskilz. I was going over this study that you linked earlier: http://journals.plos.org/plosgenetic...l.pgen.1003316

I'm struggling to understand this graph here:

Spoiler Alert, click show to read: 

So if I understand this study correctly, it's saying that minorities in the Levante cluster close to Jews, Cyptriots, and Europeans while Jordanians and Palestinians cluster close to Saudis. Wouldn't you expect Palestinians and Jordanians to cluster right between Saudis and Levantine minorities? Unless I misunderstood you, you said that Palestinians typically have 50% Levantine ancestry, and the rest being primarily Arabian. And I have absolutely no clue what the tree chart is saying.

No I think I might have a clue to what this is actually.

Have you ever read about Dienekes Womb of Nations theory? Also the West Asian and European components would maybe have some similarities to his Northern Middle Eastern population theory.

[sumskilz]

So if I understand this study correctly, it's saying that minorities in the Levante cluster close to Jews, Cyptriots, and Europeans while Jordanians and Palestinians cluster close to Saudis. Wouldn't you expect Palestinians and Jordanians to cluster right between Saudis and Levantine minorities? Unless I misunderstood you, you said that Palestinians typically have 50% Levantine ancestry, and the rest being primarily Arabian.

There are actually more Saudis stretching below what that PCA covers, so Jordanians and Palestinians are pretty much overlapping the Saudi cluster, but at the end of it that is closest to the Levantine minorities, but that still doesn't answer your question because you'd expect that Palestinians and Jordanians would be about half way between. The reason they aren't, is that all the Muslims are being pulled something like ~10% toward sub-Saharan African populations (which are very far away). The pre-Muslim conquest Levant and Arabia were basically made up of different proportions of the same three prehistoric ancestries (CHG, ENF, and Ancestral Arabia). None of those three constituent ancestries differ from each other more than Fst = 0.057. That is they were more than 94% identical at the population level in the first place. Whereas the sub-Saharan ancestries differ from those Middle Eastern ancestries by as much as Fst = 0.256. If think that if it weren't for the sub-Saharan admixture, the Saudi cluster would be much closer to the Levantine minorities cluster with the Jordanian/Palestinian cluster in between and overlapping the edges of both.

That said, the view that Palestinians have 50% Levantine ancestry is sort of a best fit for the current data based on the assumption of continuity, meaning everything that can be explained by Levantine ancestry is assumed to be Levantine ancestry, but some of it could be Levantine-like ancestry from the Hijaz. Ancient DNA will eventually sort it out better, but techniques were only recently devised for extracting a lot of autosomal DNA from bones that have been in a hot environment and it's illegal (for the most part) to excavate known burial sites in all the relevant countries except for in cases were they are about to be destroyed by construction. I don't know when the results will be published, but there were at least three Bronze Age Canaanites excavated last summer who died in a building collapse during a battle and over two hundred Iron Age Edomites from southern Jordan now in San Diego.

And I have absolutely no clue what the tree chart is saying.

It's inferring population phylogenies based on similarity, but it can't be an accurate representation because it doesn't take into account admixture. It's an interesting analysis for understanding similarities, but it actual tells you nothing about the real population history. They used fineSTRUCTURE, when I'd be more interested in TreeMix results.

No I think I might have a clue to what this is actually.

Have you ever read about Dienekes Womb of Nations theory? Also the West Asian and European components would maybe have some similarities to his Northern Middle Eastern population theory.

Yeah, it is because the West Asian and European components are so similar compared to the sub-Saharan African. Most of the Mediterranean European admixture in Jews is Levantine-like anyway, because it arrived in Europe with the Neolithic farmers. It's only the Western Hunter Gatherer component that pulls some Ashkenazi Jews off closer to Europe.

Dienekes was right about that before we had all the ancient data. Basically what is referred to as Caucasoids in the old anthropological literature really are people with varying degrees of ancestry from the early pastoralists and farmers who radiated out from the Middle East.

[Blaze86420]

Perfect, that makes total sense. Everything is starting to come together now.

[Blaze86420]

Another question if possible. How exactly does this calculator (or any other calculators and studies) figure out what admixture corresponds to which ancient populations? So for example: how do you figure out that 55% of Sardinian DNA is from the Neolithic farmers? Do they just compare DNA from broad populations against DNA from population isolates to find admixture patterns? Also, how does ancient DNA fit into any of this? Would ancient DNA serve the some purpose as population isolates, except far more accurately?

[sumskilz]

Another question if possible. How exactly does this calculator (or any other calculators and studies) figure out what admixture corresponds to which ancient populations? So for example: how do you figure out that 55% of Sardinian DNA is from the Neolithic farmers? Do they just compare DNA from broad populations against DNA from population isolates to find admixture patterns? Also, how does ancient DNA fit into any of this? Would ancient DNA serve the some purpose as population isolates, except far more accurately?

The actual mechanics and technical aspects involved require a lot of background knowledge and can only really be explained mathematically (for example), so I’ll skip to the general principles.

Programs have been developed that infer ancestry based on statistical models. At first they were tested on simple problems for which the real world answer was already known, such as identifying which individuals are African or European or Asian. As this developed, more complex questions were tested. For example, if a person has three English grandparents and one Nigerian grandparent, is the program accurately identifying their ancestry as ~75% European, ~25% African?

The most common type of admixture analysis is “unsupervised”. The autosomal genetic data for a large number of individuals is fed into the program and it is set to divide the data into a certain number of populations (K). The more data that is fed into the program, the more accurate its predictions are. Once the ancestral components have been identified, percentages can be accurately estimated for individuals. If there are sub-Saharan Africans in the samples fed in, at K=2 it will always divide sub-Saharan ancestry from the rest. At K=3, it will further divide non-sub-Saharan ancestry into East Asian/Native American versus the rest. As K increases it further subdivides groups. At first these ancestries are all phylogenetic branches of human ancestry that match those identified by physical anthropologists, but as the divisions become more detailed, it starts to identify population splits that weren’t previously known to anthropologists. Because it accurately is determining splits that are known, those that aren’t known are assumed to be real, at least a real enough possibility to warrant further investigation. Sometimes these ancestries weren’t known because nobody exists today who is primarily of that ancestry. These are called ghost populations, at least until they are identified.

There was a ghost population that was showing up as a minority percentage in Europeans. It was highest around the Baltic Sea, so it was just being called Baltic or something similar. Then ancient hunter gatherers who lived in Europe started being tested, and it turned out that they matched the Baltic ghost population. Only it wasn’t Baltic in origin, that’s just where it survived at the highest percentage. Mesolithic hunter gatherers from Iberia to Scandinavia to Greece all genetically matched this component. Previously Basques were being used as a proxy for pre-agricultural Europe because they spoke a language isolate and lived in the mountains so were at least a little bit isolated genetically, but that was based on an erroneous assumption. Basques turned out to be a partial population isolate, but not of Europe’s hunter gatherers, they are a partial isolate of the first farmers. Although not as much as the Sardinians are. Being on an island evidently causes more isolation than being in the mountains.

About Sardinians and Neolithic farmers, it's a similar story. I'll just quote Dienekes' recent post:

A couple of new papers appeared this week. First, an article in Nature on natural selection in ancient Europe includes a sample of Anatolian Neolithic farmers and concludes that the European Neolithic farmers were descended from them with a bit of extra European hunter-gatherer admixture. Second, a new preprint on the bioRxiv includes Neolithic samples from northern Greece and finds that they too resemble the Anatolian and European farmers. I think it is time to declare the problem of "Neolithization of Europe" done. It took less than 4 years to solve it with ancient DNA. Here is a (non-exhaustive) list of papers in historical review:

Keller et al. (2012): Iceman (5kya) looks Sardinian! Was this a fluke?
Skoglund et al. (2012): No, because... Swedish farmer (5kya) looked Sardinian too! When did these "Sardinians" come to Europe?
Lazaridis et al. (2014): No later than an LBK farmer from Germany (7kya) but what about western Europe?
Haak, Lazaridis et al. (2015): Spanish early farmers from northern Spain looked Sardinian too
Olalde, Schroeder et al. (2015): Ditto for Mediterranean Spain! So where did they all come from?
Mathieson et al. (2015): Anatolia!
Hofmanová, Kreutzer et al. (2015): via Greece!

Again the ancestral component had already been identified. It was just being called "Mediterranean" or "that Sardinian-like component" before anyone knew the source.

In the Levant, Druze and Samaritans make much better proxies because their respective religions have kept them from intermarrying for the last 1,000 years for Druze, and 2,500 years for Samaritans. Meaning their isolation was close to complete. So in an unsupervised admixture analysis, you can look at what ancestries make up the majority of the genome of Druze and Samaritans to get an idea of what the ancient Levant was like.

You can also do supervised admixture analysis. For example, you could say I believe Druze and Samaritans are good enough proxies for the ancient Levant that I will use their DNA to define a population component that I will call “Ancient Levant” and then based on that assumption being true, let the program estimate admixture. The weakness of doing this with proxies is that it’s based on an assumption, maybe it’s a good assumption as in this example, but still imperfect. Druze and Samaritans are small populations relatively speaking so maybe they only represent a sub-set of the ancestral population. Maybe 98% of their DNA is from the ancestral Levant but they only retain half of the diversity of what existed back then which might still exist as components of the ancestry of others. The other issue is that no matter how isolated or culturally conservative a group is, over many generations there will be at least a tiny amount inward gene flow from rape or affairs for example, so that can throw things off. Supervised admixture works better with ancient DNA because you know its source for certain. Although it’s usually just best to include the ancient DNA in the data of unsupervised analysis and see what comes out. The fact that you have an idea of what should come out just serves as a way to test the prediction.

I explained this mostly with examples, but I think that probably got the important points across. Ancient DNA helps us identify ghost populations and more accurately calculate ancestral components.

[Blaze86420]

Alright, back to your point about PCA plots. The Oracle single population sharing gives you "genetic distance" from the top 20 closest populations. Does that really mean genetic distance like you talked about being shown in the PCA plots? Or do those numbers just mean what percentage of your DNA is not typical of the corresponding population?

I hope you can understand what I'm asking.

[norse]

So are there any new studies on Western and Southwestern Norwegians? There is some R1A1 there, and lots of stocky and brown haired individuals, like myself, although I am in Minnesota descended from late 19th century and early 20th century immigrants. Carleton Coon even recognized this, and I find his writings even from way back in the early 20th century to be quite accurate. Also it seems some of this R1A1 distribution is also heavily present in Scotland.

[sumskilz]

Does that really mean genetic distance like you talked about being shown in the PCA plots? Or do those numbers just mean what percentage of your DNA is not typical of the corresponding population?

The second, it's the percentage that can't be explained well by the proposed fit. Although I guess that if you made a plot using those numbers it might look similar.

So are there any new studies on Western and Southwestern Norwegians? There is some R1A1 there, and lots of stocky and brown haired individuals, like myself, although I am in Minnesota descended from late 19th century and early 20th century immigrants. Carleton Coon even recognized this, and I find his writings even from way back in the early 20th century to be quite accurate. Also it seems some of this R1A1 distribution is also heavily present in Scotland.

This ends up more on topic here: The Archaeogenetics of Europe

[Blaze86420]

Are there any PCA plot generators out there? I want to see where I'd end up.

[sumskilz]

Are there any PCA plot generators out there? I want to see where I'd end up.

There is one on the internet, but it has some flaws because not all populations had the same markers tested, which means some populations get pulled off in weird directions a bit, like Samaritans.

Here: http://www.interpretome.com/#pca

Click "Begin Exploring" in the upper right

Click it again on the pop up

Enter the name that you want yourself labelled as (if you plan to post a screenshot)

Select the unzipped text file of your raw data

You can experiment with other settings, but the best for you will be:

Data Source: Mid-East/Jewish
Resolution 100,000
PC1
PC2

Click visualize

When it's finished, click to deselect some populations to make it less confusing. Don't expect it to be as accurate as the ones in the studies, but it'll give you some sense probably. I say probably because I know the samples they used and some individuals end up as weird outliers when that is not the case in any of the studies. If you end up in a place that makes sense, it's probably more or less accurate. There are no Palestinians in the data set, but Jordanians are pretty close.

[Blaze86420]

Don't really know what to make of it, besides that I cluster pretty close to Turks/Iranians. I'm completely outside of the Jordanian cluster, which is weird.

[sumskilz]

Don't really know what to make of it, besides that I cluster pretty close to Turks/Iranians. I'm completely outside of the Jordanian cluster, which is weird.

It makes sense that you would be pulled slightly toward Turks/Iranians, but I think it's a flaw in the design. It seems to really exaggerate some differences. See the one Jordanian and one Moroccan that are over in the middle of the Ethiopians. Them being pulled toward Ethiopians a little bit makes sense, but the degree is ridiculous. No Jordanian in that sample is more Ethiopian-like than half of Ethiopians. It seems to be giving way too much weight to particular markers.

Your PC1 position might be roughly correct (left to right on this projection), but you are probably too far toward the top on PC2 (which seems stretched in general), that's happening with all the Yemenis as well (you might have noticed), even though the Yemeni Jews are in about the right spot.

EDIT: In addition to the over-weighting of some markers that appears to be going on, I noticed that the reason everything looks stretched top to bottom (PC2) is because each axis is illustrated at a different scale. Look at the numbers, one line side to side is equal to four lines up and down. With that in mind, you are actually closer to some Jordanians than to any of the Turks. The Iranian outlier there probably atypically Arab-like (with some markers likely exaggerated in weight).

They seem to be using the same publicly available data sets as this:



Notice the Jordanian and the Iranian near each other in the lower right corner? I assume those are the two individuals nearest you. I'm guessing that without distortions, you might be near that Lebanese Muslim above them.

[Blaze86420]

I was wondering sumskilz, are there any studies done on Iraqis and Gulf Arabs?

[sumskilz]

I was wondering sumskilz, are there any studies done on Iraqis and Gulf Arabs?

Unfortunately, nobody has wanted to work in Iraq lately (for some reason), but this is interesting: Kuwaiti population subgroup of nomadic Bedouin ancestry—Whole genome sequence and analysis

It looks like they did as much as they can with a very small budget.

Also, these:

Population Genetic Structure of the People of Qatar

Inferring genome-wide patterns of admixture in Qataris using fifty-five ancestral populations

Two other bits of news that belong in this thread....

Samples from seven individuals from Wadi al-Hammeh in northern Jordan are currently in the lab at La Trobe University. They're from a Natufian culture site - literally the first farmers. Although I have no idea how long it will be until publication.

Some guy (3DRIF-26) without a head, either a gladiator or soldier, turned up in a grave from Roman Britain. Isotope analysis shows that he most likely grew up in Egypt (best fit) or the Levant (close second). I'm thinking the Levant, because of his DNA. Y-Haplogroup is J2, mtDNA is H5. He didn't cluster with any of the groups in the study, but was equally close to the Cypriot, Druze, Sardinian, Saudi, and South Italian samples.

But check this out:



He pretty much clusters with modern Samaritans. He's a 94% match for the three Samaritans from the public databases (the black circles), but even closer, except in one case, to the other Samaritans (tested at 23andMe).

[Blaze86420]

Wait, why does he cluster with Egypt at all? I thought Egypt was mostly "Back-to-Africa" at this point?

[sumskilz]

Wait, why does he cluster with Egypt at all?

He doesn't, Egypt is the best fit for the isotope analysis which isn't a genetic fit. Isotope analysis is used to determine where a person grew up because the minerals from the local environment (like from drinking water for example) end up in a person's bones. He was found in England, but the archaeologists wanted to know where he was originally from. The environment changes over time and the Levant and Egypt are very similar chemically so being a slightly better fit for Egypt isn't that strong of evidence for him having grown up in Egypt rather than the Levant. It could have been either. The Jordan Rift Valley is actually the second best fit isotopically.

His position in relation to the Samaritans is the first published ancient DNA to confirm what was assumed about them being an isolate. The Samaritans cluster really tightly near the Samaritan Levite with the exception of two outliers. That's interesting because a Levite isn't allowed to marry the descendant of a convert. The Samaritans have been endogamous for ~2500 years, and there were probably no converts after the Muslim conquest.

There is a known historical event that could explain the outliers:

“In the ninth year of Hoshea, the king of Assyria captured Samaria; he carried Israel away into Assyria” (2 Kings 17:6). Then “the king of Assyria brought people from Babylon, Cuthah, Avva, Hamath, and Sepharvaim, and placed them in the cities of Samaria instead of the children of Israel; and they took possession of Samaria, and dwelt in the cities” (2 Kings 17:24). According the Assyrian administrative records, that would have been just the ruling class, soldiers and and some craftsmen. The particular incident is confirmed by an Assyrian inscription attributed to Sargon II, “The inhabitants of Samaria, who agreed [and plotted] with a king [hostile to] me, not to do service and not to bring tribute [to Assur] and who did battle, I fought against them with the power of the great gods, my lords. I counted as spoil 27,280 people, together with their chariots, and gods, in which they trusted. I formed a unit with 200 of [their] chariots for my royal force. I settled the rest of them in the midst of Assyria. I repopulated Samaria more than before. I brought into it people from countries conquered by my hands. I appointed my eunuch as governor over them. And I counted them as Assyrians.” There have also been four cuneiform contracts found in the area which include both Hebrew and foreign names which match the ethnic identities in 2 Kings.

Judging by his position relative to the Samaritans and Arabia, my guess for the ethnic background of this gladiator or soldier is Idumean, Jewish, or Nabatean. He is pulled a bit toward Arabia from the Samaritans but both his haplogroups are more Levantine than Arabian.

[Blaze86420]

Oh woops, I somehow didn't notice the mention of isotope analysis in your first post.

[sumskilz]

There has been significant progress lately in extracting and analyzing ancient DNA. Some of this has refined our understanding and forced a bit of reinterpretation. So here is my summary of the major findings:

Prehistoric Near Eastern populations had very low Neanderthal admixture and significant Basil Eurasian ancestry. Basil Eurasians were the first out-of-Africa population to split off from the rest.

Basil Eurasian ancestry:

66% (±13) in Mesolithic Iran
48% (±6) in Neolithic Iran
44% (±8) in Epipaleolithic Levant

Basil Eurasians may not have had any Neanderthal admixture. This is confusing since the earliest Neanderthal admixture was believed to have taken place in the Middle East.

Early Neolithic Ancient Near Eastern populations were highly differentiated with population distances up to Fst=0.15 which is greater than the distance between modern Europeans and East Asians Fst=0.13. The Neolithic Levant was most distant from Neolithic Iran with Anatolia and the Caucasus being intermediate between the two.

The ancestry of the first Southern Levantine farmers was about two thirds Natufian hunter-gatherer and one third from people related to the first Anatolian farmers. The Anatolian farmers probably originated in the area of modern Southeastern Turkey and Northwestern Syria.

The first farmers in Western Iran appear to have been mostly descended from Mesolithic hunter-gatherers in the same region who were remotely related to Caucasus hunter-gatherers.

Chalcolithic people of Western Iran had a major impact on the Near East. Levantine Bronze Age population had ~44% Chalcolithic Western Iranian ancestry. Bronze Age Northwestern Anatolians had ~33% Chalcolithic Western Iranian ancestry. The Yamna culture (source of the Indo-European languages) had ~43% Chalcolithic Western Iranian ancestry.

The Neolithic Levant was the source of Neolithic admixture in North Africa and the Horn of Africa. Modern people in the Horn of Africa have ~42% Neolithic Levantine ancestry.

People of India have significant Neolithic/Chalcolithic Iranian ancestry, first from the Harrapan culture and then from the steppe. This ranges to up to ~77% in some Indian and Pakistani populations to as low as ~12% in Kharia people.

All this explains why people from India to Europe to North Africa are all much more closely related to each other than to other human populations. They are all predominately descended from Neolithic Middle Eastern populations and only to a much lesser extent from the local hunter-gatherers.

Recent publications:

The genetic structure of the world’s first farmers

Early Neolithic genomes from the eastern Fertile Crescent

The Demographic Development of the First Farmers in Anatolia

Update on the Middle Eastern gladiator from England who pretty much clustered with Samaritans. He and the Samaritans also cluster with Early Bronze Age samples from Jordan.

[Blaze86420]

Hey Sumskilz, I was looking at these two PCA plots:

Spoiler Alert, click show to read: 

They're both from the same study I believe (The genome-wide structure of the Jewish people) but they're showing different results. Any idea why?