This is a great article, I heard David had some complaints with the dating as well. I think it is also interesting how the Eneolithic steppe can be modeled with a good amount of WSHG. Would a relationship between that new TTK (Kelteminar?) sample and Steppe populations be worth looking into, as a proxy for ANE? From what I've heard, TTK is around 3/4 ANE, 1/4 Iran_N.
Also, have you any thoughts on the notion that none of the Anatolian samples have steppe DNA? I tested some of the Kaman-Kalehoyuk samples on G25 as well as qp a while back and MA2208 (low-res albeit), MA2200, and MA2203 seem to consistently have a few percentage points (0-10%) of Steppe. 2203 is probably the best bet for EHG detection in Anatolians, because it is neither low res like MA2208 (which tends to score higher EHG/Steppe) but has more than 2200.
MA2200 and 2203 are a combined sample in the reich dataset (Turkey_OldHittitePeriod.SG) and I think it scored around 5% Yamnaya on qp with a passing score (p =~0.08)
Yeah, a Siberian source does seem to be needed. TTK works, I've tried it but didn't talk about it in this article since it's not published and we don't know a lot about it. Yeah, it's a convenient proxy for ANE. I'm not the guy to ask about Anatolian samples since I don't really have much experience in modelling ancient or modern samples from there. And about those Hittite period samples, I think you should try modelling them using on rotating strategy. This way, you test all possible sources and make them compete against each other. If you're skeptical about any ancestry's presence, remove it from the sources and put it in the outgroup list and run qpAdm again, if it fails it means that the component is present and if it doesn't fail it's absent. But I myself have tested these samples like a year or two ago and I don't remember them scoring steppe ancestry... but I wasn't as good as I am now in modelling so I can't confirm either. Thanks for the comment, and are you from Anthrogenica by any chance?
Yes, I am billyh from anthrogenica. I am still pretty callow when it comes to formal stats but once I figure out rotating pops I will try that. I was remembered of this when I saw this thread so spent some of today modelling Kaman-Kalehoyuk and Ovaoren turks in G25 and qp, with varying results. Some of the models had inadequate P-values.
Both Seh Gabi C and Meshkovo passes for Progress En, but both of these populations are very distinct from each other, IIRC Seh Gabi C is like 60% Iran N and Meshkovo is like 15% and like 65% CHG i don't understand why both of them would pass considering how different they are from each other. Also how would entirely neolithic model for progress en look like?
They don't pass (p < 0.05). They're just the best models currently, as simple models consisting of EHG and CHG fail, some combo of Iran_N, ANF/PPNB and Siberian ancestry is definitely needed in some amounts for Progress_En and Vonyuchka_En. Notice the admixture percentages of Iran_C_SehGabi/Darkveti-Meshoko and CHG in both models. And now check this image https://imgur.com/a/AnNouHw, can you see that except CHG and Iran_N other components are in similar quantities? And now look at the percentages of Iran_C_SehGabi/Darkveti-Meshoko in Progress_En models, you see that CHG is higher for the SehGabi model as SehGabi lacks it and it's lesser in the Darkveti-Meshoko model as Meshoko has it's own CHG to provide. So, the proportions are all balanced.
An entirely neolithic model would be like a Khvalynsk like pop with EHG, CHG and ANE and Darkveti Meshoko. In fact we already have a proximal Neolithic model. Run the 2-way model (Khvalynsk_I0434 + Darkveti-Meshoko) for Progress individually, and you'll see 15-30% of Meshoko. This is exactly what I predict with the 4-way model in my post.
If there is Seh Gabi like ancestry then there is like 15% Iran N ancestry in Progress and if there is Meshkovo in progress then there's like 2% Iran N and plus extra 22% CHG added. That's why I said both models contradict each other, it would be model them better with only using entirely neolithic populations
I agree there is probably some other ancestry in Steppe Eneolithic than just EHG + CHG, hopefully more samples are released in the future so we can have clearer picture on the genetic origins of Steppe Populations
Here from Eurogenes. Love the post. Hope you don't mind me repeating myself a bit here.
A few thoughts:
Rather than ANE admixture in Progress, is it possible that it has a great deal of ancestry from a population that was somewhat intermediate between ANE and EHG? If our EHG "baseline" is too rich in WHG, that would obscure additional WHG contributions to Yamnaya too wouldn't it?
I think looking at the linguistic context is important, and the Kroonen paper (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0275744) is great for that. They suggest that Anatolian must have come from an early, eastern source, prior to the spread of agriculture to that region, as Anatolian languages do not share much agricultural terminologie with other IE languages, whereas "core" IE language must have come from a western, later (Yamnaya) source after the spread of agriculture. That pretty much rules out an Armenian homeland for PIE in and of itself.
I'd note that in the paper they test for Iron Gates ancestry, which I think is a good way to detect EEF ancestry.
ARM_Tavshut_Trialeti_MBA and IRN-HajjiFiruz_BA have both significant EHG and Iron Gates ancestry. I would suggest that that means these samples descend in part from western Yamnaya, and are related to "core" Indo-European speaking groups (probably proto-Armenian at least for the former).
Areni Cave Chalcolithic and the ALA026 outlier from Hatay both have EHG ancestry but no detectable Iron Gates ancestry. My suggestion is that these samples are related to Anatolian speaking populations. The Areni sample is so early too that it would give steppe ancestry plenty of time to be diluted prior to the Middle Bronze Age, much like steppe ancestry in Armenians later became diluted.
It can be possible, but we just don't have any such sample yet. You're correct, linguistics doesn't really have much going on to support this theory, Indo-Hittite was in the ruins until it was revived by Lazaridis et al. 2022. The fact that nuclear PIE has a more vast agricultural vocab than PIA which is supposed to be in proximity of the earliest region to have adopted agriculture, says a lot about this theory.
These samples (Trialeti and HajjiFiruz_BA) most likely descend from Catacomb which is genetically the same as Yamnaya.
Areni Cave has Progress-like ancestry.
Thanks for bringing this up, CupOfSoup. This confirms my claim that Yamnaya has WHG-related ancestry from Ukraine_N which itself descends from IronGates_HGs while Progress does not, since WHG signal is only detected in Trialeti/HajjiFiruz_BA and not samples from Areni Cave.
This is a great article, I heard David had some complaints with the dating as well. I think it is also interesting how the Eneolithic steppe can be modeled with a good amount of WSHG. Would a relationship between that new TTK (Kelteminar?) sample and Steppe populations be worth looking into, as a proxy for ANE? From what I've heard, TTK is around 3/4 ANE, 1/4 Iran_N.
Also, have you any thoughts on the notion that none of the Anatolian samples have steppe DNA? I tested some of the Kaman-Kalehoyuk samples on G25 as well as qp a while back and MA2208 (low-res albeit), MA2200, and MA2203 seem to consistently have a few percentage points (0-10%) of Steppe. 2203 is probably the best bet for EHG detection in Anatolians, because it is neither low res like MA2208 (which tends to score higher EHG/Steppe) but has more than 2200.
MA2200 and 2203 are a combined sample in the reich dataset (Turkey_OldHittitePeriod.SG) and I think it scored around 5% Yamnaya on qp with a passing score (p =~0.08)
Yeah, a Siberian source does seem to be needed. TTK works, I've tried it but didn't talk about it in this article since it's not published and we don't know a lot about it. Yeah, it's a convenient proxy for ANE. I'm not the guy to ask about Anatolian samples since I don't really have much experience in modelling ancient or modern samples from there. And about those Hittite period samples, I think you should try modelling them using on rotating strategy. This way, you test all possible sources and make them compete against each other. If you're skeptical about any ancestry's presence, remove it from the sources and put it in the outgroup list and run qpAdm again, if it fails it means that the component is present and if it doesn't fail it's absent. But I myself have tested these samples like a year or two ago and I don't remember them scoring steppe ancestry... but I wasn't as good as I am now in modelling so I can't confirm either. Thanks for the comment, and are you from Anthrogenica by any chance?
Yes, I am billyh from anthrogenica. I am still pretty callow when it comes to formal stats but once I figure out rotating pops I will try that. I was remembered of this when I saw this thread so spent some of today modelling Kaman-Kalehoyuk and Ovaoren turks in G25 and qp, with varying results. Some of the models had inadequate P-values.
This is complete bullshit. EHG was Turkic, Samarra Culture was Turkic, you can cry, stupid european, your stupid blog does not change the truth.
I'm not even European 💀
Both Seh Gabi C and Meshkovo passes for Progress En, but both of these populations are very distinct from each other, IIRC Seh Gabi C is like 60% Iran N and Meshkovo is like 15% and like 65% CHG i don't understand why both of them would pass considering how different they are from each other. Also how would entirely neolithic model for progress en look like?
They don't pass (p < 0.05). They're just the best models currently, as simple models consisting of EHG and CHG fail, some combo of Iran_N, ANF/PPNB and Siberian ancestry is definitely needed in some amounts for Progress_En and Vonyuchka_En. Notice the admixture percentages of Iran_C_SehGabi/Darkveti-Meshoko and CHG in both models. And now check this image https://imgur.com/a/AnNouHw, can you see that except CHG and Iran_N other components are in similar quantities? And now look at the percentages of Iran_C_SehGabi/Darkveti-Meshoko in Progress_En models, you see that CHG is higher for the SehGabi model as SehGabi lacks it and it's lesser in the Darkveti-Meshoko model as Meshoko has it's own CHG to provide. So, the proportions are all balanced.
An entirely neolithic model would be like a Khvalynsk like pop with EHG, CHG and ANE and Darkveti Meshoko. In fact we already have a proximal Neolithic model. Run the 2-way model (Khvalynsk_I0434 + Darkveti-Meshoko) for Progress individually, and you'll see 15-30% of Meshoko. This is exactly what I predict with the 4-way model in my post.
If there is Seh Gabi like ancestry then there is like 15% Iran N ancestry in Progress and if there is Meshkovo in progress then there's like 2% Iran N and plus extra 22% CHG added. That's why I said both models contradict each other, it would be model them better with only using entirely neolithic populations
I agree there is probably some other ancestry in Steppe Eneolithic than just EHG + CHG, hopefully more samples are released in the future so we can have clearer picture on the genetic origins of Steppe Populations
Here from Eurogenes. Love the post. Hope you don't mind me repeating myself a bit here.
A few thoughts:
Rather than ANE admixture in Progress, is it possible that it has a great deal of ancestry from a population that was somewhat intermediate between ANE and EHG? If our EHG "baseline" is too rich in WHG, that would obscure additional WHG contributions to Yamnaya too wouldn't it?
I think looking at the linguistic context is important, and the Kroonen paper (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0275744) is great for that. They suggest that Anatolian must have come from an early, eastern source, prior to the spread of agriculture to that region, as Anatolian languages do not share much agricultural terminologie with other IE languages, whereas "core" IE language must have come from a western, later (Yamnaya) source after the spread of agriculture. That pretty much rules out an Armenian homeland for PIE in and of itself.
I'd note that in the paper they test for Iron Gates ancestry, which I think is a good way to detect EEF ancestry.
ARM_Tavshut_Trialeti_MBA and IRN-HajjiFiruz_BA have both significant EHG and Iron Gates ancestry. I would suggest that that means these samples descend in part from western Yamnaya, and are related to "core" Indo-European speaking groups (probably proto-Armenian at least for the former).
Areni Cave Chalcolithic and the ALA026 outlier from Hatay both have EHG ancestry but no detectable Iron Gates ancestry. My suggestion is that these samples are related to Anatolian speaking populations. The Areni sample is so early too that it would give steppe ancestry plenty of time to be diluted prior to the Middle Bronze Age, much like steppe ancestry in Armenians later became diluted.
Apologies for the late reply.
It can be possible, but we just don't have any such sample yet. You're correct, linguistics doesn't really have much going on to support this theory, Indo-Hittite was in the ruins until it was revived by Lazaridis et al. 2022. The fact that nuclear PIE has a more vast agricultural vocab than PIA which is supposed to be in proximity of the earliest region to have adopted agriculture, says a lot about this theory.
These samples (Trialeti and HajjiFiruz_BA) most likely descend from Catacomb which is genetically the same as Yamnaya.
Areni Cave has Progress-like ancestry.
Thanks for bringing this up, CupOfSoup. This confirms my claim that Yamnaya has WHG-related ancestry from Ukraine_N which itself descends from IronGates_HGs while Progress does not, since WHG signal is only detected in Trialeti/HajjiFiruz_BA and not samples from Areni Cave.