Monday 2 May 2011

Hammering the iron

Trying to extend my last analysis Northwards I was faced with a few difficulties.
Finding 2nd ("African"-like) wave elements; or Aboriginal ("East Asian"-like -or perhaps Amerindian-like) in Europeans is easy, since contributor populations were quite different.
The problem lies, I think, with different subsets and possibly pre-2nd wave expansions, within what I've been calling the Mesopotamian Core. This is much harder to detect and differentiate reliably, since I don't have many if any still "unadmixed" populations from that time.
So I'll kind of have to find the unknown "metals" through comparisons of how differently mixed "alloys" relate to each other.

In supervised ADMIXTURE terms this means I generally get series of tightly related components coming from the Mesopotamian region but probably also from Eastern Anatolia and the Levant. ADMIXTURE seems to be picking a series of distinct but very close clusters in the MPC. When considering only North Africans, Basque poles stretched to become the local variety of MPC.
With more populations, there are sufficient differences that ADMIXTURE tends to fall into lumping them with aboriginal/2nd wave smaller elements into artificial components. For instance the Siberian/Amerindian-like element of some North Europeans often disappears into the local MPC; the Nile Core of Southerners and especially Middle Easterners into their local varieties. Thus in many runs "Basque" becomes dominant in Europeans, erasing most of Siberian, "Sandawe" expands to the point of having Mesopotamian Core-like fsts incompatible with North East African origins, and ends up including most Middle Easterner diversity. South Europeans end a mix of mostly "Basque" with some "Sandawe5". The "Sandawe5" in a few cases swaps places with the "Dogon5" pole everything else similar, proving IMO that these components are real and not directly related (although perhaps somewhat distantly akin) to actual Dogon or Sandawe. And the whole thing is often very "chunky".
Unsupervised ADMIXTURE on the other hand always picks out supposedly but in my opinion not generally "unadmixed" group of individuals (see Irula in the "Indian cline"). It kind of makes it's own poles up. If original populations are not present in the data, results are not reliable (in terms of actual progenitor populations, they're reliable for relationships if properly interpreted).

Neolithic Revolutions are quite uncompromising after a "critical point" in their maturation. Before they reach this point however, they consist of distinct if related populations or subsets competing with each other, and perhaps mixing and differentiating constantly via short breath regional expansions.

The MPC seems to be composed of such a group of subsets. Here are some I think may be hidding in the data:
1) A "Basque centred" Med-Atlantic Wave subset important from Italy to the British Isles and beyond, but also present in Central Europe. Probably from Western Anatolia via a maritime coastal village-to-village route. Very similar to another Western Anatolian wave via the river valleys of central Europe.
2) A "Chuvash centred" East wave subset predominant in the Northeast but present in much of Europe. Probably from the Caucasus and Northeast Anatolia via the Steppe river valleys. It may have later spread with new agricultural developments for colder environments or poorer soils (Rye).
3) A Levant-East Anatolian element, now centred in Armenians, Druze, Georgians. I'm not sure if this isn't the parent component to the Basque and Chuvash ones, thus being distinct due to greater diversity. Meaning it could be a first level MPC population and the Basque and Chuvash ones being peripheral subsets of this subset. It could be also, or simultaneously, an "inner-core" MPC element that superseded the others in a posterior expansion. Y-haplogroup markers seem to support this possibility. This subset seems to be the one which made it all the way into India. It was probably present in Mesopotamia as well.

As I said all these components are very closely related. That's why in unsupervised ADMIXTURE you get apparently surprising results, such as the Northeast Europe-modal component being closest to the Anatolian/Levantine one (makes sense if you realise one is a subset of the other).

I increasingly think it's likely none of the above subsets emerged in an unadmixed form, in any population, from the settling of the Neolithic Revolution. Some seem to have admixed with elements from other non-MPC cores. Others mixed with other subsets of the same MPC core (for instance in the Corded Ware area). Still others with possible forager elements. Middle Eastern populations all have this presumably Northeast African element I've named the Nile Core. From preliminary runs, I feel that some European populations are less admixed than others, but all seem to have significant admixture from other elements. Which creates a problem concerning appropriate pole populations.

I will dedicate the next few days to search for different "Mesopotamian Core"-only subsets and expansions. For lack of suitable public sets, I will include participants data, but please don't read too much into the first experimental provisional results. When I have a model that makes more sense, it will be apparent I think.

3 comments:

  1. I do not fully understand what you mean by this.

    ''With more populations, there are sufficient differences that ADMIXTURE tends to fall into lumping them with aboriginal/2nd wave smaller elements into artificial components. For instance the Siberian/Amerindian-like element of some North Europeans often disappears into the local MPC; the Nile Core of Southerners and especially Middle Easterners into their local varieties''

    Do you mean that the Siberian and Amerindian like components in some Europeans are subsumed or ''swallowed up'' by the larger MPC components?

    Are the Siberian and Amerindian components that have been shown in previous runs in many Europeans still real then?

    ReplyDelete
  2. I'm very confident they're real, although absolute percentages may not be exact to the decimal digit. I'm not questioning previous results just pointing out that the situation becomes more complex the more closely I look.
    "Macroscopically", if you run Europeans as Fertile Crescent versus Amerindian/Siberian there is little complication and results are straightforward I think.
    The problems start when you want to differentiate between different "Mesopotamian" waves. And see the North African components, the Amerindian ones, and "Mesopotamian" ones all at once.
    Different "Mesopotamian" subsets are admixed with each of these more exotic elements, and when they split, they tend to take some of it with them, or to become subsumed into them.
    In order to differentiate more precisely different "Mesopotamian" and other elements in Europeans I would ideally use unadmixed modern populations corresponding to the ancient subsets. But these mostly don't exist anymore. For instance French Basques are almost fully "Mesopotamian core". But I now am seeing results suggesting even they have are composed of one dominant and another smaller "Mesopotamian" subset.
    Things will become clearer with the next few posts.

    ReplyDelete
  3. Ahhhh alright I understand now. This actually makes sense and thank you for clearing this up and answering my questions.

    ReplyDelete