An attempt at deep West Eurasian phylogeny

[Horatio McCallister]

I think this discrepancy comes from trying to force the idea of East/West Eurasian that's based on modern people on these ancient samples, Kostenki was chosen as West because it's the first ancient European that's closer to modern Europeans (who mostly come from recent Middle Easterners, and likely not at all from Kostenki) than Asians. Then as WHG is more Western, they are the "pure" West, then the Middle East... we stop there, they are Basal instead?

I've been thinking that the split between West and East Eurasians is a lot fuzzier than just a clean bifurcation at some point in time X number of years ago. Following from this thought, would you say the earliest Bacho Kiro samples really do represent an early Asian population as Kale and the authors of that original paper suggested? Or rather, could it just be that latter Asians descend from a BK/IUP population, probably from Central Asia, and that BK only has an affinity to East over West Eurasians because later WEs simply have more layers of various "Basal"-esque ancestries which dilutes their affinity to BK/IUP compared to East Eurasians? Given that the earliest diffusion of IUP/OoA had to have occurred somewhere out of Southwest Asia, I find it very hard to believe that the direction of population movement really went from Bacho Kiro to East Asia, as opposed to somewhere in Central Asia to East Asia.

[MrI1]

If modern Haplogroup I diverged apart from the Pre-Haplogroup I ancient Vestonice Cluster samples from Northeastern Austria and Southeastern Czech Republic around 35,000 years ago, what are the odds that both cousin branches would be found only in Europe if this patrilineal divergence occurred outside of Europe and could be explained by separate migrations into Europe? All things being equal wouldn’t the simplest explanation be that this common patrilineal ancestor lived in Eastern or Southeastern Europe around 35,000 years ago?

[Woz]

If modern Haplogroup I diverged apart from the Pre-Haplogroup I ancient Vestonice Cluster samples from Northeastern Austria and Southeastern Czech Republic around 35,000 years ago, what are the odds that both cousin branches would be found only in Europe if this patrilineal divergence occurred outside of Europe and could be explained by separate migrations into Europe? All things being equal wouldn’t the simplest explanation be that this common patrilineal ancestor lived in Eastern or Southeastern Europe around 35,000 years ago?

That would make more sense, but there's not much room in Europe for unadmixed WHG's to hide in. Eastern Europe was Gravettian-like (Sunghir, Kostenki, Buran Kaya), and Gravettians seem to be a mixture of WHG and Ust Ishim/Yana_UP, whereas the Epigravettians seem to be 'pure' WHG's.

Epigravettian probably started in, and spread from the Balkans, which is where we should be looking for unadmixed WHG's, but the Balkans were also Gravettian like (Pestera Muierii in Romania, Bacho Kiro in Bulgaria) before the LGM.

[kolompar]

I think this discrepancy comes from trying to force the idea of East/West Eurasian that's based on modern people on these ancient samples, Kostenki was chosen as West because it's the first ancient European that's closer to modern Europeans (who mostly come from recent Middle Easterners, and likely not at all from Kostenki) than Asians. Then as WHG is more Western, they are the "pure" West, then the Middle East... we stop there, they are Basal instead?

I've been thinking that the split between West and East Eurasians is a lot fuzzier than just a clean bifurcation at some point in time X number of years ago. Following from this thought, would you say the earliest Bacho Kiro samples really do represent an early Asian population as Kale and the authors of that original paper suggested? Or rather, could it just be that latter Asians descend from a BK/IUP population, probably from Central Asia, and that BK only has an affinity to East over West Eurasians because later WEs simply have more layers of various "Basal"-esque ancestries which dilutes their affinity to BK/IUP compared to East Eurasians? Given that the earliest diffusion of IUP/OoA had to have occurred somewhere out of Southwest Asia, I find it very hard to believe that the direction of population movement really went from Bacho Kiro to East Asia, as opposed to somewhere in Central Asia to East Asia.

Yes, simple f4s show them clearly on the Eastern side, and I'm thinking exactly the same on how it's possible with their age. The Bacho Kiro paper has a graph on their mtDNA haplogroups that's very similar to the Y-haplogroup case, they belong to early splitting M and N haplogroups next to all kinds of Asians on the tree, while "West Eurasians" are later subhaplogroups of N (with BK-1653 being U). Maybe "East Eurasian" is just the default unadmixed OoA?

[ilabv]

I think this discrepancy comes from trying to force the idea of East/West Eurasian that's based on modern people on these ancient samples, Kostenki was chosen as West because it's the first ancient European that's closer to modern Europeans (who mostly come from recent Middle Easterners, and likely not at all from Kostenki) than Asians. Then as WHG is more Western, they are the "pure" West, then the Middle East... we stop there, they are Basal instead?

I've been thinking that the split between West and East Eurasians is a lot fuzzier than just a clean bifurcation at some point in time X number of years ago. Following from this thought, would you say the earliest Bacho Kiro samples really do represent an early Asian population as Kale and the authors of that original paper suggested? Or rather, could it just be that latter Asians descend from a BK/IUP population, probably from Central Asia, and that BK only has an affinity to East over West Eurasians because later WEs simply have more layers of various "Basal"-esque ancestries which dilutes their affinity to BK/IUP compared to East Eurasians? Given that the earliest diffusion of IUP/OoA had to have occurred somewhere out of Southwest Asia, I find it very hard to believe that the direction of population movement really went from Bacho Kiro to East Asia, as opposed to somewhere in Central Asia to East Asia.

Yes, simple f4s show them clearly on the Eastern side, and I'm thinking exactly the same on how it's possible with their age. The Bacho Kiro paper has a graph on their mtDNA haplogroups that's very similar to the Y-haplogroup case, they belong to early splitting M and N haplogroups next to all kinds of Asians on the tree, while "West Eurasians" are later subhaplogroups of N (with BK-1653 being U). Maybe "East Eurasian" is just the default unadmixed OoA?

Isn't it more sensible that Zlaty Kun, Ust Ishim, etc are OoA (plus some Neanderthal)

[Enki]

I think this discrepancy comes from trying to force the idea of East/West Eurasian that's based on modern people on these ancient samples, Kostenki was chosen as West because it's the first ancient European that's closer to modern Europeans (who mostly come from recent Middle Easterners, and likely not at all from Kostenki) than Asians. Then as WHG is more Western, they are the "pure" West, then the Middle East... we stop there, they are Basal instead?

I've been thinking that the split between West and East Eurasians is a lot fuzzier than just a clean bifurcation at some point in time X number of years ago. Following from this thought, would you say the earliest Bacho Kiro samples really do represent an early Asian population as Kale and the authors of that original paper suggested? Or rather, could it just be that latter Asians descend from a BK/IUP population, probably from Central Asia, and that BK only has an affinity to East over West Eurasians because later WEs simply have more layers of various "Basal"-esque ancestries which dilutes their affinity to BK/IUP compared to East Eurasians? Given that the earliest diffusion of IUP/OoA had to have occurred somewhere out of Southwest Asia, I find it very hard to believe that the direction of population movement really went from Bacho Kiro to East Asia, as opposed to somewhere in Central Asia to East Asia.

Yes, simple f4s show them clearly on the Eastern side, and I'm thinking exactly the same on how it's possible with their age. The Bacho Kiro paper has a graph on their mtDNA haplogroups that's very similar to the Y-haplogroup case, they belong to early splitting M and N haplogroups next to all kinds of Asians on the tree, while "West Eurasians" are later subhaplogroups of N (with BK-1653 being U). Maybe "East Eurasian" is just the default unadmixed OoA?

one of the Yana samples had U, and they were very early.

[Desdonas]

@all

I am confused about the mysterious affinity between West Eurasian and core East Asian populations, especially the direction of geneflow. Could someone make an attempt below, and if it is significant, then a small WE to Core EA geneflow may be possible. Otherwise, the Core-EA to later WE direction would be a better choice.

f4(Mbuti, Kostenki; Tianyuan/AR33K, Primorsky/Boshan/Liangdao)

In addition, I heard in another paper that Goyet Q116-1 has a special relationship with Tianyuan which does not even exist in AR33K. Try this f4 below would help to confirm it.

https://www.science.org/doi/10.1126/science.abi8202#F2

f4(Mbuti, Goyet Q116-1; AR33K/Onge/Hoabinhian, Tianyuan)

[Norfern-Ostrobothnian]

Chimp.REF Russia_Kostenki14 China_Tianyuan Taiwan_Liangdao_LatePaleolithic.SG -0.000361 0.00116 -0.310 0.756 118965
Chimp.REF Russia_Kostenki14 China_Tianyuan Russia_DevilsCave_N.SG 0.00157 0.000631 2.49 0.0129 843345
Chimp.REF Russia_Kostenki14 China_Tianyuan Laos_Hoabinhian.SG 0.000340 0.000770 0.441 0.659 414712

Kostenki at least has no preference over the other except Devils Cave which probably has ANE ancestry. And I don't think West Eurasians are actually a mixture of an East Eurasian related and less East Eurasian related branch as a whole. I think GoyetQ116-1 is mixed but not West Eurasians as a whole

Chimp.REF Belgium_UP_GoyetQ116_1 China_Tianyuan Laos_Hoabinhian.SG -0.00224 0.000964 -2.32 0.0203 224458
Chimp.REF China_Tianyuan Russia_Kostenki14 Belgium_UP_GoyetQ116_1 0.00319 0.000862 3.70 0.000214 467391
Chimp.REF China_Tianyuan Russia_Kostenki14 Romania_PesteraMuierii_EUP 0.000409 0.000756 0.541 0.589 843738

[Kale]

I am confused about the mysterious affinity between West Eurasian and core East Asian populations, especially the direction of geneflow. Could someone make an attempt below, and if it is significant, then a small WE to Core EA geneflow may be possible. Otherwise, the Core-EA to later WE direction would be a better choice.

I just tried a run, and successfully managed to resolve the WE-EA connection by separating the East-Asian > Basal mixture into...
1) Onge > Basal mixture
2) Early ANE (Yana-like) > East-Asian mixture.
It's an extra admixture above the OP graph, but it's more logical and a better statistical fit, perhaps I was just being too conservative with the admixture count.

Also...
I've been thinking of a way to make the coastal origin of East Eurasians feasible from what we see autosomally.
Maybe something like this could work...
Stage 1: Levant
Stage 2: Zagros
Stage 3: Ust-Ishim's ancestors cross the Northern Zagros through one of the valleys, follow the East Caspian up into the steppe, spreading both East and West
Stage 4: East-Eurasian proper's ancestors follow the Zagros South to the coast and spread from there, maybe until they hit the plains in Northern China?
Stage 5: East-Eurasians spread West across the steppe absorbing increasing levels of Ust-Ishim ancestry as they go, Tianyuan/AR33k end up with ~1/3, BachoKiro probably closer to 2/3, and Goyet would be almost entirely Ust-Ishim.

[Desdonas]

I am confused about the mysterious affinity between West Eurasian and core East Asian populations, especially the direction of geneflow. Could someone make an attempt below, and if it is significant, then a small WE to Core EA geneflow may be possible. Otherwise, the Core-EA to later WE direction would be a better choice.

I just tried a run, and successfully managed to resolve the WE-EA connection by separating the East-Asian > Basal mixture into...
1) Onge > Basal mixture
2) Early ANE (Yana-like) > East-Asian mixture.
It's an extra admixture above the OP graph, but it's more logical and a better statistical fit, perhaps I was just being too conservative with the admixture count.

Also...
I've been thinking of a way to make the coastal origin of East Eurasians feasible from what we see autosomally.
Maybe something like this could work...
Stage 1: Levant
Stage 2: Zagros
Stage 3: Ust-Ishim's ancestors cross the Northern Zagros through one of the valleys, follow the East Caspian up into the steppe, spreading both East and West
Stage 4: East-Eurasian proper's ancestors follow the Zagros South to the coast and spread from there, maybe until they hit the plains in Northern China?
Stage 5: East-Eurasians spread West across the steppe absorbing increasing levels of Ust-Ishim ancestry as they go, Tianyuan/AR33k end up with ~1/3, BachoKiro probably closer to 2/3, and Goyet would be almost entirely Ust-Ishim.

OK. But one problem is that, according to the f4 results above, Liangdao and Tianyuan have the same Kostenki affinity. So the ANE to East Asia geneflow is specific for ANEA, but not for proto-EA or ASEA?

Also, Tianyuan had already obtained Denisova ancestry, which is absent in early European samples including Bacho Kiro (if they derive 33% ancestry from E/SE Asia). So a North India-Central Asia route for BK (split before East Asians and Onge/Hoabinhian) is more parsimonious, and Tianyuan is an admixture of an early EA and a BK-related lineage which spread east.

Last, I would move the location of EE-proper/Ust split, as well as the origin of K2 more east, probably in Sistan. This could explain why the Eurasian expansion during 50-40kya is strongly associated with K2, and the lack of K2 in Ancient/Modern Near East (except definite ANE derived geneflow).

[CJC]

I am confused about the mysterious affinity between West Eurasian and core East Asian populations, especially the direction of geneflow. Could someone make an attempt below, and if it is significant, then a small WE to Core EA geneflow may be possible. Otherwise, the Core-EA to later WE direction would be a better choice.

I just tried a run, and successfully managed to resolve the WE-EA connection by separating the East-Asian > Basal mixture into...
1) Onge > Basal mixture
2) Early ANE (Yana-like) > East-Asian mixture.
It's an extra admixture above the OP graph, but it's more logical and a better statistical fit, perhaps I was just being too conservative with the admixture count.

Also...
I've been thinking of a way to make the coastal origin of East Eurasians feasible from what we see autosomally.
Maybe something like this could work...
Stage 1: Levant
Stage 2: Zagros
Stage 3: Ust-Ishim's ancestors cross the Northern Zagros through one of the valleys, follow the East Caspian up into the steppe, spreading both East and West
Stage 4: East-Eurasian proper's ancestors follow the Zagros South to the coast and spread from there, maybe until they hit the plains in Northern China?
Stage 5: East-Eurasians spread West across the steppe absorbing increasing levels of Ust-Ishim ancestry as they go, Tianyuan/AR33k end up with ~1/3, BachoKiro probably closer to 2/3, and Goyet would be almost entirely Ust-Ishim.

Archaeologically, East Eurasian ancestral population probably spread from the steppe, specifically the Tianyuan-related branch has material evidence connecting it to Upper Paleolithic Altai industries about 40.000 years ago.

I find your model about Yana and East Asians interesting, but are Northern East Asian groups such as Primorsky or DevilsGateCave_N closer to Yana than Tianyuan to Yana? Or closer to Yana than Onge? Usually they appear equal related in stats I have seen.

However, Yang et al. (2020) in the paper "Ancient DNA indicates human population shifts and admixture in northern and southern China" (also co-authored by Fu Qiaomei) suggest some presence of ANE ancestry in northern East Asian populations such as Yumin or Boshan in their qpAdm runs (proxied by Kolyma-related "Paleosiberian" ancestry). Perhaps this evidence for the Yana-related gene flow.

[Kale]

Liangdao is a low coverage noisy sample, I wouldn't read too much into it.
The Denisova discrepency is something I hadn't considered though, good point.
Do you know/remember which populations the D0, D1, D2 Denisovan lines where present in?

[Desdonas]

Liangdao is a low coverage noisy sample, I wouldn't read too much into it.
The Denisova discrepency is something I hadn't considered though, good point.
Do you know/remember which populations the D0, D1, D2 Denisovan lines where present in?

IIRC

Only D0 in Tianyuan, Salkhit and Yana.

Only D2 in Australians, and Papuans also have some local D1. Both no D0 (almost no segment overlapping Salkhit's Denisova ancestry).

D2-like in South Asians (AASI) ,no D0.

Both D2 and D0 in East/Southeast Asians.

[Kale]

Archaeologically, East Eurasian ancestral population probably spread from the steppe, specifically the Tianyuan-related branch has material evidence connecting it to Upper Paleolithic Altai industries about 40.000 years ago.

I find your model about Yana and East Asians interesting, but are Northern East Asian groups such as Primorsky or DevilsGateCave_N closer to Yana than Tianyuan to Yana? Or closer to Yana than Onge? Usually they appear equal related in stats I have seen.

However, Yang et al. (2020) in the paper "Ancient DNA indicates human population shifts and admixture in northern and southern China" (also co-authored by Fu Qiaomei) suggest some presence of ANE ancestry in northern East Asian populations such as Yumin or Boshan in their qpAdm runs (proxied by Kolyma-related "Paleosiberian" ancestry). Perhaps this evidence for the Yana-related gene flow.

Yana and Salkhit (who is Tianyuan-heavy in ancestry) seem to have some bidrectional flow. You can see MA1 prefers the East-Asians over China_UP, but Yana doesn't. That is because China_UP is closer to Yana, not that East-Asians are closer to MA1.
Mbuti.DG Yana_UP.SG China_UP Jomon 0.00024 0.53 918598
Mbuti.DG Yana_UP.SG China_UP RUS_Primorsky_Boisman_MN 0.00000 -0.01 962533
Mbuti.DG Yana_UP.SG China_UP Taiwan_Hanben_IA -0.00035 -0.96 1028933
Mbuti.DG MA1.SG China_UP Jomon 0.00144 2.57 661818
Mbuti.DG MA1.SG China_UP RUS_Primorsky_Boisman_MN 0.00127 2.60 690766
Mbuti.DG MA1.SG China_UP Taiwan_Hanben_IA 0.00120 2.61 733082
Mbuti.DG China_UP Yana_UP.SG MA1.SG -0.00139 -2.67 734005
Mbuti.DG Onge.DG Yana_UP.SG MA1.SG 0.00007 0.13 787129
Mbuti.DG Jomon Yana_UP.SG MA1.SG 0.00001 0.03 683539
Mbuti.DG RUS_Primorsky_Boisman_MN Yana_UP.SG MA1.SG 0.00006 0.12 721979
Mbuti.DG Taiwan_Hanben_IA Yana_UP.SG MA1.SG 0.00022 0.50 799607

IIRC

Only D0 in Tianyuan, Salkhit and Yana.

Only D2 in Australians, and Papuans also have some local D1. Both no D0 (almost no segment overlapping Salkhit's Denisova ancestry).

D2-like in South Asians (AASI) ,no D0.

Both D2 and D0 in East/Southeast Asians.

That helps resolve the love triangle between Onge/Tianyuan/East-Asians nicely. Without considering Denisova there are 3 possibilities...
1) East-Asian = Onge + Tianyuan
2) Onge = East-Asian + X (something more generic Eurasian)
3) Tianyuan = East-Asian + X
Denisova seems to constrain things to option 1.

EDIT: Also I'm inches from creating a working 'CWE graph'.
- Goyet is the 'archetypal West Eurasian' (AWE)
- Then there is 'archetypal West-Asian' (AWA) (Basal + Southern-ENA mix), Iran_N has the most of that, 66% AWA, 22% CWE, 12% ANE
0) A 72/28 mix of AWE and AWA happens.
1) First to branch off is 'CWE'
2) Next to branch off is ANE's Western portion
3) Then comes a big portion of Muierii/BK1653, they mix 50/50 with something not particularly Goyet-related, but Goyet-like in that it missed the AWA mixture. Then they each get ~20% CWE.
4) Kostenki branches off
5) Sunghir finishes that tree
6) Gravettians = BK1653 + Sunghir as usual
7) WHG = 71% CWE, 23% Gravettian, 6% ANE
8) Georgia_UP/Pinarbasi = 73% CWE, 27% AWA
9) Taforalt = 60% ANA, 24% CWE, 16% Pinarbasi

With that setup, CWE is probably some sort of South/Southwest Balkan thing, and likely also responsible for Levantine Aurignacian.
The only real thing left to take care of is a few 0 edges in the AWA realm, but I don't think that will be terribly difficult to resolve.

[Song]

We dont know what kind of Denisovan paleo ancient samples carry on mainland asia, only that it's not that the one Papuan/Australians encountered on their island.

Denisovan segment sharing with ancient DNA: 

Asians: Cambodians 2/38, Tu 4/21, Koreans 2/31, Japanese 0/50, Yi 0/47. - Japanese and Yi don't carry those segments for some reason.

Siberians : Eskimo 0/30, Yukat 0/36, Chukchi 0/9, 1/18 Itelman, Tubalar 2/39 - Less common among Siberians for some reason.

Americans: Mayan 0/34, Mixe 0/53, Karitiana 1/42. - Mayan and Mixe don't seem to carry those segments for some reason.



Higher segment sharing with some East Asians and Southeast Asians, but less common in Siberians.