Experimental K12 "Deep" Calculator for Africans and West Eurasians

[Aben Aboo]

No distance basic model

Distance: 2.8362% / 0.02836154
57.0 POL_Globular_Amphora
27.6 Yamnaya_RUS_Samara
6.6 TUR_Barcin_N
4.0 Ancestral_N_African
2.6 Caspian
0.8 Basal_West_African
0.6 Pre-Pastoralist
0.4 Basal_Central-West_African
0.4 Malaysia_LN_2550BP

[TuaMan]

Recently I posted a speculative Out of America hypothesis here. Part of what intrigued me about this hypothesis was a recent paper by Max Planck & Ian Mathieson which posited a Eurasian back-migration into Africa. This paper has been received otherwise on this forum with confusion, since it doesn't seem to "make sense" with the traditional phylogeny... I will repost a quick excerpt here:



To be clear, this calculator doesn't "prove" and isn't meant to prove the broader Out of America theory I posted. The calculator's purpose is to show that Africans can be modeled with much more Upper Paleolithic Eurasian ancestry than previously considered-- which opens up the way to some type of pre-Neolithic back-migration.

Have you tried running any qpGraphs or other tests with the Bacho Kiro samples to see if they also work in modeling Africans?

[K33]

I just tried a few...

In this graph you can see Somali take 42% ancestry from a ghost clade that contributes most of Bacho Kiro's ancestry, with the other 58% from something a little more basal than Taforalt.

But most of that is coming from Levant_N, which is by process of elimination b/c Levant_N is the least basal sample for Bacho Kiro to draw from out of this set. So perhaps this result is not extremely informative or surprising...



The issue I guess is that we lack adequate "bridge" samples, we have neither unadmixed Basal Eurasians nor unadmixed Basal Africans. And it becomes more difficult to fit the tree adding additional African moderns too... when I added Yoruba they took ~25% from Shum Laka, with the rest of their ancestry being somewhat similar to the mix Somali got (lost that graph, the score was 150+ though, not quite as good)

[Squad]

Quick rundown of the leaderboard for the main components:

Taforalt is 53% ANA. Compare to the Shum Laka paper (54% Ghost North African) and the original Taforalt paper (45% ANA)... pretty good. (The Yamnaya in Taforalt is overfitting, but for the Berbers and maybe even the Fulani it may be real, if perhaps slightly inflated):



Ancient Arabian was created to improve the fits for Gulf Arabs and to provide better proxies for the Arabian ancestry in Habesha. Within this component should probably be something like ~20-35% Basal Eurasian ancestry:



The Pre-Pastoralist component. As stated elsewhere, this component seems to have substantial West Eurasian ancestry "locked inside" of it, but it's hard to pull it out considering how many variables there are in East Africa. It's stands about midway between the Basal_East_African and the Ancient_Arabian components, in terms of single-item distance.



The Central-East component was designed around the ~27% Eurasian back-migration ancestry Mathieson et al attributed to Mbuti. The remainder forms the basis of the Central-East African component. It was impossible to get substantial "West African" ancestry to go into Mbuti, as modeled in the Shum Laka paper. They need their own component to make the fits work-- not just for them but for other ancients and moderns with residual "pygmy" ancestry.



I didn't want to make a second "pygmy" component, but as predicted by the Shum Laka paper, where the Central African lineage splits very early into "East" and "West" branches, this is necessary. It was impossible to get fits for Biaka below 0.23 (!) using the Central_East component that fit so well for Mbuti-- even in combination with the West and East African components. So the Shum Laka paper appears to be on to something here with the deep basal split within the rainforest. Note also the significant role played by these Central African lineages in the ancient Malawi samples.



The Basal East African component was modeled mostly on the Nilo-Saharan-like Kenya_Kakapel_900BP sample, and I used the back-migration paper's attribution of ~41% Eurasian ("Ust Ishim") ancestry for them. There is some minor spillover from Ust Ishim into other Eurasian components but overall it works. This is also satisfying since, among moderns, the Mursi show the highest Basal_East_African ancestry even thought I didn't specifically target them-- Mursi have been cited in other academic papers, admixture graphs, etc, as exemplifying the "Nilo-Saharan" component. Dinka, OTOH, have a slight West African shift; Mota (ETH_4000BP) has a South-Central African shift.



For the West Africans, I originally targeted Yoruba then adjusted this with Lemande/Esan. I had Yoruba at ~35% Ust Ishim/65% Basal West, but then as per the Taforalt paper, which attributed ~13% Taforalt-related ancestry into Yoruba, I "backed out" ANA from the West African component. Since about half of Taforalt's ancestry is ANA (the rest Eurasian) you can see the results here. The Barcin and Villabruna-like ancestry in West Africans are thus mediated mostly thru Taforalt. Gambians have more ANA ancestry than Yoruba, which makes sense. Note ALSO, how Mende have a nice little chunk of Central African East/West ancestry? The two Central African components are the most "basal" of all those in this set. It makes sense then that the Mende pick this up-- the Shum Laka paper attributed to them slightly more of the deep "Ghost Modern" and "Ghost Archaic" ancestry than Yoruba...



Finally the South African component. I tried to make this peak in South Africa_2000BP (ZAF2000), then add back ~15% of some type of Pastoralist ancestry into JuHoan as per the literature. But it just doesn't work. It's not even about the Ust Ishim component, even if you remove this, the JuHoan are more "distal" from Eurasians than ZAF_2000BP. You can see this on a PCA too. Instead, ZAF pull in some of the Central_East/West components. Does the Basal_South_African contain one or more of these rainforest components nested within it-- and ZAF_2000BP simply have an excess of it relative to JuHoan?

What is most interesting in your runs is how Arabians mostly lack Barcin, this is even reflected in East Africans as well of course. Obviously we would expect Arabians to have a fairly decent amount of ENF ancestry from the levant if P58 came from there. The Caspian component is as usual high in Arabians, almost as high as in the Levant. Could this indicate that my theory about P58 migrating from Mesopotamia/Iran directly into Arabia can indeed be correct? Or is it only that the ancient arabian you created just happens to be a highly drifted component which results in it "eating up" barcin-like ancestry among bottlenecked populations, with Yemenis scoring some barcin because their less bottlenecked ancestry allows it to pop out? Anyways, I've always suspected something was off for many reasons. This is an interesting experiment.

[K33]

What is most interesting in your runs is how Arabians mostly lack Barcin, this is even reflected in East Africans as well of course. Obviously we would expect Arabians to have a fairly decent amount of ENF ancestry from the levant if P58 came from there. The Caspian component is as usual high in Arabians, almost as high as in the Levant. Could this indicate that my theory about P58 migrating from Mesopotamia/Iran directly into Arabia can indeed be correct? Or is it only that the ancient arabian you created just happens to be a highly drifted component which results in it "eating up" barcin-like ancestry among bottlenecked populations, with Yemenis scoring some barcin because their less bottlenecked ancestry allows it to pop out? Anyways, I've always suspected something was off for many reasons. This is an interesting experiment.

It may eat a little bit, but this relative lack of Barcin-related ancestry holds up even when using Natufian. The dividing line is where the Fertile Crescent ends and the desert begins in the South Levant. Palestinians (Christians, Samaritans) and Oriental Jews have much, much more Barcin-related acnestry than Sinai Bedouin or Arabians.

The Arabian component I created is basically just a "southern-shifted" Natufian, it's only 0.04% away from Natufian in G25 IIRC, but its shifted *away* from Barcin. (ie, "Ancient Arabian" is designed to have slightly more Basal Eurasian ancestry than Natufian)

[Magnetic]

any update ?