Thursday, May 24, 2018

What's Maykop (or Iran) got to do with it?


I had a go at imitating this qpGraph tree, from the recent Wang et al. preprint on the genetic prehistory of the Caucasus, using the ancient samples that were available to me. I'm very happy with the outcome, because everything makes good sense, more or less. The real populations and singleton individuals, ten in all, are marked in red. The rest of the labels refer to groups inferred from the data.


However, this is still a work in progress, and, if possible, I'd like simplify the model and also get the worst Z score much closer to zero. If anyone wants to help out, the graph file is available HERE. Feel free to post your own versions in the comments, and I'll run them for you as soon as I can.

Update 31/05/2018: I've managed to come up with a new version of my model that not only offers a better statistical fit, but, in my opinion, also a much more sensible solution. For instance, the Eastern Hunter-Gatherer node now shows 73% MA1-related admixture, which, I'd say, makes more sense than the 10% in the previous version. The relevant graph file is available here.


For more details and a discussion about the updated model, including additional trees with Baltic Corded Ware and British Beaker samples, please check out my new thread on the topic at the link below.

What's Maykop (or Iran) got to do with it? #2

Citation...

Wang et al., The genetic prehistory of the Greater Caucasus, bioRxiv, posted May 16, 2018, doi: https://doi.org/10.1101/322347

See also...

Yamnaya: home-grown

Ahead of the pack

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

38 comments:

  1. Seems very complex--maybe too much so. You also have ALOT of 0 edges--which suggests that the connections may not be quite correct.

    I think one of the issues maybe that ANE--at least as MA-1 as an example--really look like a admixture event--A beringia population (EA related) mixing with a WHG-like population. I think your graphs with a EA or SA anchor looked better than these--it brought in that something extra into the EHG/ANE group

    Also the Yamnya group does not seem connected to CHG? they both share a BE connection--but should it not be a bit more direct than that? The EHG component seems Really High for the Steppe Eneolithic and for Khvalynsk.

    I might play with this a bit this weekend

    ReplyDelete
  2. Blogger Lee Albee said...
    "Seems very complex--maybe too much so. You also have ALOT of 0 edges--which suggests that the connections may not be quite correct........."


    What a great idea. Nobody has created such a schematic.
    Parsing the data to exclude outliers; adding more detailed information [like specific snp's per cluster ] can't hurt either.

    ReplyDelete
  3. @Lee Albee

    Seems very complex--maybe too much so.

    I was trying to copy this tree. The complexity here is "hidden" by the different type of graphical presentation.

    https://2.bp.blogspot.com/-5lUVvfzgbng/Wv5s6R1z4OI/AAAAAAAAGys/gPPvcQJdTE0UPsROEGtgPba8mg19NYSvQCLcBGAs/s1200/Wang_etal_admix_graph.png

    Also the Yamnya group does not seem connected to CHG? they both share a BE connection--but should it not be a bit more direct than that?

    That's how Wang et al. chose to design their tree for Yamnaya_Caucasus, and I wanted to see if it also works for Yamnaya_Samara.

    The EHG component seems Really High for the Steppe Eneolithic and for Khvalynsk.

    If the "basal" admixture in the Eneolithic populations on the Steppe wasn't really from CHG, but a more basal, closely related population, then that can explain the higher EHG.

    ReplyDelete
  4. @Davidski

    Oh your model is better than the wang one. F4 distance is better than wang.

    The edges still look bad. Their bases assumption to have MA1 without admixture is probably biting them.

    ANE feeding into MA1 should be mixed. EHG is likely additional admixture into the ANE population of more WHG like ancestry or more ANE as AG2 and AG3 seem to be better fits as a source for ANE into native americans than MA1.

    ReplyDelete
  5. So Wang et al supposes that there was little gene flow through the Caucasus because Maikop or related don't appear to be a donor to Steppe groups, and as they see it the uniformly different y-chromosomes points to a lack of gene flow between North and South.

    Rather, they view the supposed 40% CHG-related ancestry in steppe groups as mostly just a dash of BE into the bigger EHG soup that creates a mirage of Caucasus ancestry in steppe groups. They probably have a few horse-eaters east of the seas with additional BE that might support their view.

    If I'm not totally misinterpreting the issue, here's the problem I see with that. It can be shown anthropological, culturally and linguistically that a lot of the stuff on the steppe came from the south, probably the west too after this. They're trying to solve the foreign brides dilemma with a phantom population solution that makes no sense in any case, mind mtdna.

    Maybe that's a stupid comment, but I'm having a hard time with this.




    ReplyDelete
  6. @Lee Albee
    Looks like Native Americans are better modeled with their ANE as AG3 while EHG is better modeled as WSHG+WHG (tail prob better than with or including AG3) in qpAdm. MA-1 is very old so not surprising.

    ReplyDelete
  7. @bellbeakerblogger

    The Eneolithic Steppe samples already resemble Yamnaya very closely, rather than Maykop, and they're older than anything to do with Maykop.

    So now, genetically, Yamnaya can't be derived from Maykop or Steppe Maykop, not to any significant degree. It's impossible, especially since Steppe Maykop is really weird and unexpected, and probably part Botai.

    Yamnaya looks like a population that formed via the sex biased mixture between two distinct hunter-gatherer groups in Ciscaucasia - Eastern Euro Hunter-Gatherers and a population very similar to Caucasus Hunter-Gatherers - plus admixture, again largely via female-mediated gene flow, from European farmers/herders like Globular Amphora, Trypilla, or Baden.

    It's likely that there was continuous, mostly female-mediated gene flow into Yamnaya from the Caucasus and surrounds, but it wasn't huge, so if Maykop contributed anything to Yamnaya, it was mainly just culture and, well, maybe language.

    But if we ignore the somewhat strange conclusions in the Wang et al. preprint, and focus just on the authors' main findings, that is also very difficult to accept, because the authors point out that there was something of a genetic and cultural border between the ancient Steppe and Caucasus peoples, which shifted back and forth, until finally Maykop disappeared and Yamnaya took over the North Caucasus steppes.

    This looks to me like they competed with each other, and may have regularly been in conflict, rather than swapped language and poetry. What do you think?

    By the way, Maykop is sometimes credited with being the source of Kurgans and wagons in Yamnaya. But aren't the Kurgans and wagons much older in the Balkans and Central Europe than in the Caucasus? If so, and considering the western, Globular Amphora-related admixture in Yamnaya (16% in my and Wang's qpGraph trees!), isn't it possible that Yamnaya got its Kurgans and wagons from west of the Steppe?

    ReplyDelete
  8. @Davidski,

    Ok, I think I'm tracking. Thanks for clarifying. I always thought Maikop was too late, too convenient. However I do think the Southern ancestry is a valid and real thing in PIE, whether Kartvelian-Euphratean, Proto-East Semetic, East Anatolian. I think early PIE needs these influences. But as you suggest these mixed marriages probably happened over a long period of time.

    My personal view is that large work horses and wagons originate in Europe with MLN groups such as Lengyl, TRB and GAC. That's where the tangible evidence is.

    ReplyDelete
  9. David,

    Here is a theory to explain the Iran_N/CHG in Eneolithic steppe and steppe Maykop without Anatolian_N admixture. What if the Iran_N/CHG is coming from across the Caspian Sea from Central Asia where we had populations with minimal to non-existent ANF admixture such as Namazga_CA, Sarazm_EN or Dali_EBA ? A migration from that distance will also explain the East Asian admixture in steppe Maykop.

    ReplyDelete
  10. Ukraine I3719 elite EEF male taking EHG wives
    Just like north Caucasus CHG taking EHG wives
    See the pattern ?
    And no, Myceneans don’t have 30% steppe you joker

    ReplyDelete
  11. @Jaydeepsinh Rathod

    What if the Iran_N/CHG is coming from across the Caspian Sea from Central Asia where we had populations with minimal to non-existent ANF admixture such as Namazga_CA, Sarazm_EN or Dali_EBA?

    That's impossible because Yamnaya, Corded Ware and other closely related groups don't have any of that type of ancestry, which is admixed with something related to the Botai people.

    Genome-wide and uniparental marker data show this clearly, so I'm not sure why you're suggesting it?

    Steppe Maykop does have that type of ancestry, but as I pointed out in my comment above, it's not ancestral to Yamnaya, Corded Ware etc. Not even partly so.

    Steppe Maykop and Yamnaya are distinct groups, and eventually Yamnaya takes over from Steppe Maykop in the steppes north of the Caucasus.

    ReplyDelete
  12. @Rob

    Quit acting crazy.

    There's no Y-hg J in any of the Eneolithic/Bronze Age steppe groups, so it's extremely unlikely that it was common practice for Caucasus males to take steppe wives and settle on the steppe, although maybe it happened from time to time?

    And I never said Mycenaeans had 30% steppe admixture. My qpGraph and qpAdm models show a maximum of just over 20% Steppe_MLBA-like admixture in them.

    ReplyDelete
  13. @Rob

    In this tree, which of the populations are rich in R1b?

    https://2.bp.blogspot.com/-5lUVvfzgbng/Wv5s6R1z4OI/AAAAAAAAGys/gPPvcQJdTE0UPsROEGtgPba8mg19NYSvQCLcBGAs/s1200/Wang_etal_admix_graph.png

    CHG-rich CHG and Maykop?

    or

    EHG-rich EHG and Eneolithic_Steppe?

    Simple question. Can you answer it?

    ReplyDelete
  14. Archaeology seems to suggest a link between the Zarzian Culture of the Southern Caspian and the Yangelskaya Culture of the Southern Urals. Are there any link between Yangelskaya and the Botai Culture ? Maybe the type of CHG that mostly mixed into EHG was from the Southern Urals ?

    When they say more Basal then certainly we have to look at a population nearer to the time of origin of CHG which would be +-14 000 years ago ? So CHG formed after Haplogroup R migrated from Siberia to the West. Did R evacuate from the area between the Don, Urals and Caucasus completely and some Basal CHG people spread into this area just before Haplogroup R returned from the West during the Later Mesolithic ?

    At the end of the day only more Mesolithic and Late Upper Palaeolithic samples will tell....

    ReplyDelete
  15. @ Davidski

    Maybe in the Ice Age, the north Caucasus experienced female-mediated CHG introgression (ie the females of J populations from the south moving into an R1 north). In fact, this isn't the case because we have J in mesolithic Russia already.
    But as far as what happens from the 5th millenium onwards, that is not particularly relevant although highly interesting (& still rather speculative).
    The events is becoming clear from the cumulative lines of evidence - a privileged group (of mostly R1b males with laudable CHG, nut perhaps also Q1) spread throughout the steppe, and further into central Eurasia and Europe, most likely emanating from the north Caucasus piedmont steppe, not surprising given the concentration of 'chiefly' wagon burials exactly at the Kuban steppe in the later, Yamnaya period. See Fig 2

    These groups spread CHG (along with EHG) into areas previously lacking it (Samara, Ukraine), not by women being ferried through it, like some commentators hope to suggest.

    You also continue to be wrong about the nature of EEF introgression on the other side of the steppe. Take a look at the I2a2a1b males which seem to appear during the Mariupol phase, who although highly variable, all seem to have EEF, and one man (I3719) is 100% EEF. As I'm sure you're aware - females don't carry Y haplogroups (although they maybe do in this post-modern world), but i have no doubt that females also were responsible for EEF introgression.

    Now someone like yourself should by now be able to grasp these basic tenets, so the only reason you seem intent on promulgating errors is something personal to you and the narrative you wish to create for a segment of readers. Personally, Im neither hg J nor R1b nor Georgian nor a female, so have no horses in this race.

    ReplyDelete
  16. @Rob

    Eneolithic/Bronze Age steppe populations are largely a mixture of EHG and CHG.

    Their Y-chromosomes are dominated by R1, which is a North Eurasian lineage, while their mtDNA haplotypes are often phylogenetically rooted in the Near East.

    So if they don't derive from populations mostly made up of Eastern European males and, in large part, Near Eastern females, then this is all a remarkable coincidence.

    I don't believe in such remarkable coincidences, and I expect that all of the data will back me up in the end.

    I also expect that when you're finally proven wrong on this, just like you were about Maykop being responsible for the southern admixture in Yamnaya, Corded Ware + R1a-M417 not being from the steppe, etc. you'll find something else to argue about, but I'm getting tired of these debates.

    ReplyDelete

  17. Dave


    "By the way, Maykop is sometimes credited with being the source of Kurgans and wagons in Yamnaya. But aren't the Kurgans and wagons much older in the Balkans and Central Europe than in the Caucasus? If so, and considering the western, Globular Amphora-related admixture in Yamnaya (16% in my and Wang's qpGraph trees!), isn't it possible that Yamnaya got its Kurgans and wagons from west of the Steppe?"


    Better late than never!

    ReplyDelete
  18. @Rob: While your explanation is possible, it is still far less likely and all the evidence points to steppe males. You just speculate or which facts support your hypothesis?

    ReplyDelete
  19. @ Dave
    Yes it’s hard to perfectly guess where exactly where the CHG was from. Contrary to your imperfect reading , I did not suggest any necessarily one to one relation to Majkop
    Quite the contrary I expected north Caucasus to harbour some rather northern groups and admixture to have entailed multiple events
    So, I was certainly closer than your brides theory.
    And yes , R1a is from the forest steppe primarily
    Sorry if that doesn’t suit your fantasies

    @ Zardos
    Yes . Have a read of the recent paper on the Caucasus that was published. It should be understandable, even for you

    ReplyDelete
  20. How many R1b samples from CHG heavy individuals do you have so far?

    ReplyDelete
  21. ...something of a genetic and cultural border between the ancient Steppe and Caucasus peoples, which shifted back and forth, until finally Maykop disappeared and Yamnaya took over the North Caucasus steppes. This looks to me like they competed with each other, and may have regularly been in conflict, rather than swapped language and poetry

    Climatic reasons ?

    ReplyDelete
  22. One thing that seems pretty unsatisfying about both Wang and Jeong's paper (more so Wang's paper) is that we don't really have any attempts to measure continuity or discontinuity, between them, with modern populations in the Caucasus. Jeong attempts to quantify ancestry in today's Caucasus populations without access to any actual Caucasus ancient samples, while Wang's paper simply does not have any modelling of this sort (to my recollection).

    This is actually pretty important as indicative, at least circumstantial, evidence for our understanding of how early the Caucasian languages got to where they are, and whether the Caucasus ultimately provided a barrier for the spread of any language.

    Previously, I've given a lot of credence to the suggestion of spread of today's Caucasian language families into the Caucasus in LBA or early Iron Age. That is what some linguists believe the divergence times would allow and suggest.

    That allows for either a migration from the Caucasus to the steppe, or vice versa, without the presence of Caucasian languages proving a barrier (because they were not there yet).

    But great genetic continuity in the Caucasus may instead add some support theories that Caucasian languages diversified in the Caucasus at an early date, and that IE languages may never have flowed through the Caucasus. Or at least past a very early date.

    ReplyDelete
  23. @Rob

    “And yes , R1a is from the forest steppe primarily”

    Yes, and this is the only link between Europe and India, genetically, linguistically, culturally, i.e. the only haplogroup that was truly Indo-European.
    They should stop using the term Indo-European while referring to other haplogroups, languages and religions. Other names should be introduced for example for populations from Caucasus which influenced steppe and which you so wrongly call Indo-European.

    ReplyDelete
  24. @ Zardos

    Happy to answer , although I answered your question already -
    PG2004, PG2001, VJ1001 from this study - R1b and probably 50-60% CHG looking at their admixture bar. Not to mentioned the Z2103's throughout Yamnaya with ~ 40% 'CHG'. That's 'CHG rich' is it not ?
    There's also I1635 (L388 xM269) from Kalavan Cave, Kura Araxes , Armenia c. 3000 BC, ~ 70% CHG.
    So plenty of R1b in/ from "CHG rich' areas.
    Hope this sets you on a path to competence.


    @ EastPole

    I don't know exact details, but I agree overall. There's no reason to assume that everyone on or near the steppe spoke the same language. Som groups might have spoken more along the Caucasus lines. Linguists had also linked Basque to this area.

    ReplyDelete
  25. @Rob: But how can you know that R1b came from the CHG side in this samples, if R1b seems to be older in Eastern Europe than the Caucasus and the typical CHG yDNA haplogroups are others obviously?
    Why should exactly that group on the fringes of the CHG territory be the only carrier of R1b in the CHG spectrum?
    Also, Yamnaya/steppe seems to be more in line with a Southern female mtDNA variation. How come that they are predominantely R1b, predominantely mtDNA wise Southern, but still more EHG than CHG?
    I see the possibility that your explanation is right, but it is not more probable than the steppe male hypothesis with the data available until now. You have to close a lot of gaps for that and I think the CHG male hypothesis will fail for the steppe.
    Don't say I know it, but neither do you. Therefore its not justified to attack people proposing the steppe male hypothesis the way you did.

    ReplyDelete
  26. @ Matt

    "But great genetic continuity in the Caucasus may instead add some support theories that Caucasian languages diversified in the Caucasus at an early date, and that IE languages may never have flowed through the Caucasus. Or at least past a very early date."

    Perhaps when looking a global population genomic data alone.
    However, we also need to consider later/ ongoing movements north t south and vice -versa. E.g. after 3000 BC there is a movement of K-A culture to NE Caucasus.
    Abkhazians in NW Caucasus are heavily G2 (compared to the predominantly J2 and L) Majkop. The North Caucasus Dolmen culture is considered a separate entity to Majkop.
    Lastly, we actually need to dig up the pre 4500 BC situation in the piedmont.

    ReplyDelete
  27. @ zardos

    Meh you're simply not getting it/ refusing to understand it. You're thinking in binary terms.

    ReplyDelete

  28. Blogger a said...
    zardos

    @ Zardos

    PG2004, PG2001[5304-6012YBP+/-], VJ1001 are clustered under Steppe in PCA/Admix.
    VJ1001 can't qualify as R1b Yamnaya as the sample is a female.
    The authors of the paper place the above samples with sample SA6004 ydna Q- quite Steppe like.

    ReplyDelete
  29. R1b p297 and r1b v88 have been found in Mesolithic Europe. That's all the evidence needed to make a strong argument r1b m269 originated in Europe.

    Wait till we get a decent collection of ehg y DNA. R1b m269 will pop up.

    ReplyDelete
  30. @Le skipper de Pytheas

    Climatic reasons?

    Very likely, but people usually don't just leave their homes and the land of their ancestors, whom they honored with giant Kurgans and whatnot, just because of gradual climatic shifts and, say, lower crop yields.

    What often happens is that they're weakened by such factors, and then either forced out or wiped out by competing groups better suited to the new conditions.

    ReplyDelete
  31. @All

    I updated the (rather messy) tree from my first attempt. I got rid of the unnecessary ghost populations and most of the zero drift edges. The new version is, I think, much cleaner and sensible, and also has a little better worst Z score.

    Check it out and let me know what you think. However, keep in mind that this isn't the final version. I'm still looking to improve the worst Z score, so if anyone has ideas how to do that please don't be shy.

    By the way, the original tree and graph file are still available at the same links here.

    https://drive.google.com/file/d/1lwpjMsk_FhAvzXBwmC6Fz99vAejBtTtG/view?usp=sharing

    https://drive.google.com/file/d/1i-RLD9HIwcAAane--Lx7lb_wfWvycV08/view?usp=sharing

    ReplyDelete
  32. @Matt
    In the past, there was a tendency to think that in the areas such as Caucasus and Altai the big variety of languages must have a common origin and there must have been an initial unity and diversification happened with time. However, lexical evidence using the classical method of regular sound changes has not been able to prove the common origin adequately. Differences in syntax and morphology are significant and the internal evolution does not support an earlier greater similarity between e.g. Altaic languages. Therefore, I think that the languages in Caucasus and in the old Altaic family may have followed the opposite path of their earlier different identity and a later convergence due to their ending up in the area geographically and culturally connected. In general, the linguists have emphasized divergence and reconstructed trees based on linear separations. However, convergence has probably been a very important factor in the shaping of language families.

    ReplyDelete
  33. @Rob, I don't really know if autosomal population data between Maykop and present day Caucasus (esp. haplotypes) really does indicate a lot of continuity or not, I'm just saying it would be a good piece of additional evidence to actually test. Of course it's not definitive and if uniparental lineages indicate discontinuity (without this being attributable to drift and founder effects), then this is relevant evidence too.

    @Kristina, I guess I'm really thinking about divergences and convergences *within* each Caucasian language family (Kartvelian, NE Caucasian, NW Caucasian). I'm not sure whether we're on the same page or you're discussing a common origin for these families (which from my knowledge you are right is very debatable).

    I think you're correct to note that contact effects can deflate divergence estimates (languages in continuous contact may have begun diverging a longer time ago with contact suppressing rate of divergence from each other and maybe even the common ancestor) between even related languages, and this makes language families confined to a small area hard to compare with far flung families.

    ReplyDelete
  34. Just read an interesting book about Anatolia. It seems that the idea of Hittites being a numerically small very recently arrived elite isn't well founded. From 'Ancient Turkey.."

    "The greatest number, not quite a majority, were Hittite, with much smaller numbers of those that could be understood as Luwian. Some belonged to the unaffiliated Hurrian language, but these individuals appear to have come from the south, and if the Hurrians had a homeland on the plateau, it was in eastern rather than central Anatolia. Another unaffiliated language is one we call Hattic or Hattian, which gave its name to the land of Hatti before the Hittites came into prominence. Garelli was puzzled that so few names of this type were represented—fewer than 40, and some of these may actually have belonged to Hittites—suggesting that perhaps the language had already died out. Other names in the tablets may be nicknames or shortened forms which cannot be affiliated with anything from the information we have. 36 On the basis of their personal names, it would appear that the majority of the population at Kültepe spoke an Indo-European language of the Anatolian family. This is a crucial fact for both earlier and later Anatolian history."..

    ReplyDelete
  35. @Arza

    I really think you need to sort out some issues. You'd be a very valuable commentator here if you decided to do so.

    Case in point, this model has no purpose, and is basically infeasible.

    left pops:
    Latvia_MN_I4435
    West_Siberia_N_I5766
    Europe_LNBA_RISE568
    CHG

    numsnps used: 27222
    chisq: 7.761 tail: 0.73457122
    best coefficients: 0.173 1.218 -0.391


    ReplyDelete
  36. @Rob
    I suppose this is about tablets found at Kültepe? Because at Kültepe, that "the greatest number, not quite a majority, were Hittite" and "that the majority of the population at Kültepe spoke an Indo-European language of the Anatolian family" is exactly what european Hittitologists are saying for years now.

    But Kültepe is neither Anatolia, nor Hatti. That Hittites (who condidered Kültepe/Kanes as their original place) were a large minority, but not the majority at Kanes, is saying many things. And among them the fact they were intrusive in Anatolia. And that they were likely a far more reduced minority in other Hattian cities.

    Someday, you should read Freu/Mazoyer and the sort.

    You seem to discover something which is consensual for decades.

    ReplyDelete
  37. @Davidski

    I tried a rather simplistic gpGraph strategy to model Steppe_EMBA without CHG

    I used:
    population: 0 Mbuti 10
    population: 1 Han 43
    population: 2 WHG 5
    population: 3 Natufian 6
    population: 4 EHG 3
    population: 5 Steppe_EMBA 26

    My worst stats were:

    worst f-stat: Han WHG EHG Ste -0.008152 -0.008789 -0.000637 0.001476 -0.431


    the graph can be seen here:
    https://drive.google.com/open?id=1fBkupeOADGMj20Cp7zuxJWbkj2mH-fZp

    That part is relatively simple

    Swapping out Natufian with Anatolia_N get fairly similar score.

    It does get messier after that though when you try and add in additional populations


    ReplyDelete
  38. @Lee Albee

    Thanks. I'm a big fan of simple trees like that.

    ReplyDelete

Read the rules before posting.

Comments by people with the nick "Unknown" are no longer allowed.

See also...


New rules for comments

Banned commentators list