Uralic populations in the qpAdm models by Zeng et al. 2023

[Jaska]

Zeng et al. write: “A genetic turnover by ~4.5kya saw the emergence of a population in Northeast Siberia, Yakutia_LNBA. Today, this ancestry tends to be the only East Asian ancestry present among Uralic-speaking populations, a striking feature not shared by any other ethnolinguistic grouping.”
https://www.biorxiv.org/content/10.1101/...1.560332v1

I collected all plausible “double true” qpAdm models (TRUE/TOSI in both: 1. no negative components; 2. P > 0.05 unless no model has that) for the Uralic populations from Zeng et al. 2023, Supplementary file S6. Eastern ancestry here means Central, East and Northeast Asian ancestries, including the Yakutia, Tyumen, Mongolia, Altai, China, and BMAC ancestries.

As far as I can see, their qpAdm results do not justify their interpretation that we could now conclusively exclude all the other eastern ancestries except the Yakutia ancestry from the Uralic ethnogenesis. And certainly we cannot exclude the European ancestries (CWC related and EEF related).


The Estonians:
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.03
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02

The Finns:
Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.02
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.01

The Karelians:
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.04
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.03
Altai_N + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.02
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.01

The Vepsians:
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.04
Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.03
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA+Russia_Srubnaya = 0.01

The Saamis:
Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.97
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.93
BMAC + Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.90
Germany_EN_LBK + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.90
BMAC + Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.80
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.79
BMAC + Germany_EN_LBK + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.71
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.63
Tyumen_HG + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.63
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.46
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.43
Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.13
Altai_N + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.09
Altai_N + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.09
EHG + Yakutia_LNBA + Russia_Srubnaya = 0.06
Tyumen_HG + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.06
Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.05

The Mordovians:
Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.08
Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.08
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.05
Tyumen_HG + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.05

The Maris:
Yakutia_LNBA + Russia_Srubnaya = 0.53
Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.49
Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.46
BMAC + Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.46
EHG + Yakutia_LNBA + Russia_Srubnaya = 0.45
Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.42
Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.41
Tyumen_HG + Yakutia_LNBA + Russia_Srubnaya = 0.41
Tyumen_HG + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.41
Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.37
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.36
Germany_EN_LBK + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.36
Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.34
Tyumen_HG + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.33
EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.33
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.33
Altai_N + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.32
Altai_N + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.32
Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.29
Tyumen_HG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.29
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.29
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.28
BMAC + Germany_EN_LBK + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.28
Tyumen_HG + BMAC + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.27
Hungary_EN_HG_Koros + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.25
BMAC + Germany_EN_LBK + EHG + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.25
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.23
Altai_N + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.22
Tyumen_HG + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.22
Altai_N + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.21
Tyumen_HG + BMAC + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.21
Germany_EN_LBK + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.20
Altai_N + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.19
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.18
Tyumen_HG + BMAC + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.15
Altai_N + Germany_EN_LBK + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.12
Altai_N + BMAC + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.11
EHG + Mongolia_N_North + Russia_Srubnaya = 0.06
Germany_EN_LBK + EHG + Mongolia_N_North + Russia_Srubnaya = 0.06
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + Russia_Srubnaya = 0.05

The Komis:
Germany_EN_LBK+EHG+Yakutia_LNBA+Russia_Srubnaya = 0.005
Tyumen_HG+Germany_EN_LBK+Yakutia_LNBA+Russia_Srubnaya = 0.001

The Udmurts:
EHG + Yakutia_LNBA + Russia_Srubnaya = 0.05
BMAC + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.03
EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.03
Tyumen_HG + Yakutia_LNBA + Russia_Srubnaya = 0.02
Tyumen_HG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.02
BMAC + Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.02
BMAC + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.01

The Mansis:
BMAC + Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.76
BMAC + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.61
EHG + Yakutia_LNBA + Russia_Srubnaya = 0.55
Tyumen_HG + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.47
BMAC + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.44
Altai_N + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.42
Tyumen_HG + Yakutia_LNBA + Russia_Srubnaya = 0.41
EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.41
Tyumen_HG + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.37
Tyumen_HG + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.34
Tyumen_HG + Germany_EN_LBK + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.32
Tyumen_HG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.29
Altai_N + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.29
Tyumen_HG + BMAC + Germany_EN_LBK + Yakutia_LNBA + Russia_Srubnaya = 0.29
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.26
Tyumen_HG + Germany_EN_LBK + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.20
Tyumen_HG + BMAC + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.17

The Khantys:
EHG + Yakutia_LNBA + Russia_Srubnaya = 0.01
Tyumen_HG + Yakutia_LNBA + Russia_Srubnaya = 0.01
BMAC + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.01

The Selkups:
EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.36
Altai_N+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.34
Tyumen_HG+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.29
BMAC+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.26
EHG+Mongolia_N_North+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.23
Altai_N+Tyumen_HG+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.23
Altai_N+BMAC+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.19
EHG+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.18
Tyumen_HG+Germany_EN_LBK+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.17
Tyumen_HG+Germany_EN_LBK+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.16
Altai_N+EHG+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.14
BMAC+EHG+Mongolia_N_North+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.14
BMAC+EHG+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.13
Tyumen_HG+EHG+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.12
EHG+Yakutia_LNBA+Russia_Srubnaya = 0.10
Altai_N+EHG+Yakutia_LNBA+Russia_Srubnaya = 0.11
Altai_N+BMAC+Germany_EN_LBK+EHG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.11
BMAC+EHG+Yakutia_LNBA+Russia_Srubnaya = 0.09
Altai_N+BMAC+EHG+Yakutia_LNBA+Russia_Srubnaya = 0.07
Altai_N+BMAC+EHG+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.07
Tyumen_HG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.06
Altai_N+Germany_EN_LBK+EHG+Yakutia_LNBA+Russia_Srubnaya = 0.06

The Enets:
EHG + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.47
EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.46
EHG + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.44
Tyumen_HG + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.33
Tyumen_HG + EHG + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.32
BMAC + EHG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.29
BMAC + EHG + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.29
BMAC + EHG + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.27
Tyumen_HG + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.25
Tyumen_HG + Hungary_EN_HG_Koros + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.22
Tyumen_HG + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.20
BMAC + Hungary_EN_HG_Koros + EHG + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.17
Tyumen_HG + Hungary_EN_HG_Koros + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.17
Tyumen_HG + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.16
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + Yakutia_LNBA + Russia_Srubnaya = 0.16
Tyumen_HG + Hungary_EN_HG_Koros + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.16
Tyumen_HG + Germany_EN_LBK + Mongolia_N_North + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.11
Tyumen_HG + Germany_EN_LBK + China_YR_MN + Yakutia_LNBA + Russia_Srubnaya = 0.09

The Nganasan:
Tyumen_HG+Mongolia_N_North+Yakutia_LNBA+Russia_Srubnaya = 0.03
Tyumen_HG+Mongolia_N_North+China_YR_MN+Yakutia_LNBA+Russia_Srubnaya = 0.02

[Psynome]

I think the word "tends" is doing most of the work in that statement by the authors. As you pointed out and plenty of others have shown, there are definitely other East Asian ancestry sources in a number of Uralic speaking groups. While for reasons of a preponderance of evidence I believe Yakutia_LNBA is the best candidate for a Pre-Uralic source population, there isn't and probably will never be enough evidence to make it conclusive in the same sense as linguistic classification.

[Kale]

What are the right pops?

[ANIEXCAVATOR]

Zeng et al. write: “A genetic turnover by ~4.5kya saw the emergence of a population in Northeast Siberia, Yakutia_LNBA. Today, this ancestry tends to be the only East Asian ancestry present among Uralic-speaking populations, a striking feature not shared by any other ethnolinguistic grouping.”
https://www.biorxiv.org/content/10.1101/...1.560332v1

I collected all plausible “double true” qpAdm models (TRUE/TOSI in both: 1. no negative components; 2. P > 0.05 unless no model has that) for the Uralic populations from Zeng et al. 2023, Supplementary file S6. Eastern ancestry here means Central, East and Northeast Asian ancestries, including the Yakutia, Tyumen, Mongolia, Altai, China, and BMAC ancestries.

As far as I can see, their qpAdm results do not justify their interpretation that we could now conclusively exclude all the other eastern ancestries except the Yakutia ancestry from the Uralic ethnogenesis. And certainly we cannot exclude the European ancestries (CWC related and EEF related).

Its very important to read the paper carefully to understand precisely what the authors meant there when they made that statement. When drawing the connection between Yakutia_LNBA and Uralic populations, the authors highlighted two sides to the relationship:

1. Yakutia_LNBA is the only East Asian ancestry found among all Uralic populations, without exception.
2. Yakutia_LNBA explains all of the East Asian ancestry found among Uralic populations.

When modeling the Uralic populations using qpAdm, the authors chose to present the simplest model that passes (i.e., the one with the smallest number of source populations) as the one closest to the truth, even if it might not capture the *entire* truth or the *most detailed* truth. This choice has to be made because there is no simple way to compare qpAdm models with different numbers of source populations. Yes, more complex models have higher p-value, but more complex models are also less parsimonious and under some contrived circumstances, at least in theory, you can keep increasing p-values by continuously increasing the number of sources.

Restricting the comparison to the simplest models that pass for each population, the first statement--that Yakutia_LNBA is present in all Uralic populations and is required for them to pass--is unequivocably true, and the models you posted actually show that. This continues to be the case even when you look at more complex models.

The second statement is actually *also* true, if you look at the simplest models only. If you look at more complex models, then the second statement only tends to be true. 

This is what is indicated by the two orange dots on top of the qpAdm models in the figures showing qpAdm output for each Uralic population. Almost all Uralic populations have two orange dots above them, while other ehtnolinguistic groups typically have nothing.

About whether or not Srubnaya (i.e. Western Steppe European-like ancestry) is also present among all Uralic populations, the models the authors presented do not show that it is present in all Uralic populations--because Srubnaya is forced to be present in every single model, it is never an option for it to not be among the sources. Srubnaya as a generic European-like source is present in every single model for all admixed Inner Eurasian populations, to soak up the West Eurasian ancestry, because the authors were more interested in the East Asian ancestries of these Eurasian popualtions. The % of Srubnaya is very small in Nganasan and within two standard errors, indicating Nganasan has no significantly detectable western steppe ancestry. Kale, if you have the time, could you also replicate the Nganasan model without Srubnaya? What is needed here is to show that the model for Nganasan fails if Srubnaya is moved to the outgroups, which I think it won't.

[Kale]

I only have 1240k moderns in my dataset, Nganasan are HO panel only.

[Jaska]

Its very important to read the paper carefully to understand precisely what the authors meant there when they made that statement. When drawing the connection between Yakutia_LNBA and Uralic populations, the authors highlighted two sides to the relationship:

1. Yakutia_LNBA is the only East Asian ancestry found among all Uralic populations, without exception.
2. Yakutia_LNBA explains all of the East Asian ancestry found among Uralic populations.

When modeling the Uralic populations using qpAdm, the authors chose to present the simplest model that passes (i.e., the one with the smallest number of source populations) as the one closest to the truth, even if it might not capture the *entire* truth or the *most detailed* truth. This choice has to be made because there is no simple way to compare qpAdm models with different numbers of source populations. Yes, more complex models have higher p-value, but more complex models are also less parsimonious and under some contrived circumstances, at least in theory, you can keep increasing p-values by continuously increasing the number of sources.

Thank you. So, they made choices in interpretation of the data, and then they propose their own choice as the result? Sounds methodologically suspicious: "We aim for X, therefore we make choices toward X". No wonder the result is X, then.

I think the recent results concerning the Indo-Europeans show that we should not aim for the most simple explanation, because reality is rarely so economical. Unfortunately we still have no ancient DNA samples from the region and the time relevant for Proto-Uralic, so indirect approach is all we have.

Restricting the comparison to the simplest models that pass for each population, the first statement--that Yakutia_LNBA is present in all Uralic populations and is required for them to pass--is unequivocably true, and the models you posted actually show that. This continues to be the case even when you look at more complex models.

The second statement is actually *also* true, if you look at the simplest models only. If you look at more complex models, then the second statement only tends to be true.

I would say only that it could be true, but the model containing only the Yakutia ancestry is not considerably stronger than other models, based on their numeric values. That has been my point. And when the aim is to trace the language, the framework of the linguistic results could give us hints for choosing the model, instead of only aiming toward simplicity.

For example, it is known that the Proto-Samoyedic language spread from the south the north. Its homeland is located north from the Sayan Mountains, because PSy had contacts with Tocharian, Iranian, and two primary branches of Turkic. Keeping this in mind, should we not lean towards models showing also both the Mongolia and the Tyumen ancestries in the Selkups, the Enets and the Nganasans? (Quite high in the list for them all.) Interestingly, the Altai ancestry only shows in the Selkups, perhaps because they have assimilated the Yeniseian speakers.

About whether or not Srubnaya (i.e. Western Steppe European-like ancestry) is also present among all Uralic populations, the models the authors presented do not show that it is present in all Uralic populations--because Srubnaya is forced to be present in every single model, it is never an option for it to not be among the sources. Srubnaya as a generic European-like source is present in every single model for all admixed Inner Eurasian populations, to soak up the West Eurasian ancestry, because the authors were more interested in the East Asian ancestries of these Eurasian popualtions. The % of Srubnaya is very small in Nganasan and within two standard errors, indicating Nganasan has no significantly detectable western steppe ancestry.

OK, they did not compare different steppe-related ancestries, so we cannot know whether it is actually the same steppe ancestry in all the Uralic populations. Hopefully somebody sometimes studies also this topic.

Nganasans differ even from the other Samoyedic populations, being very similar to the Tundra Yukaghirs and the Dolgans (both non-Uralic populations). They were also the northernmost vanguard of the Samoyedic expansion, and the only population moving clearly beyond Yenisei to Central Siberia. Therefore it is not reasonable to consider the Nganasans as a proxy for the Proto-Samoyedic population, and much less for the Proto-Uralic population. For this same reason, the possible lack of the steppe ancestry in them is not a counter-argument against the presence of the steppe ancestry in the Proto-Uralic population.

[Parastais]

Did they find “Jakutia” above noise level in modern Hungarians and Estonians?
Sorry, didn’t read the article.

[Queequeg]

Did they find “Jakutia” above noise level in modern Hungarians and Estonians?
Sorry, didn’t read the article.

As you can see the model with Yakutia_LNBA is not very good for Estonians, however by removing EHG it only gets worse and the same is true if you replace EHG with Tyumen_HG.

The Estonians:
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.03
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02

[Jaska]

Did they find “Jakutia” above noise level in modern Hungarians and Estonians?
Sorry, didn’t read the article.

For some reason they did not include the Hungarians in their calculations. One might suspect that was because the modern Hungarians would not show any traces of the Yakutia ancestry, thus nullifying their pre-decided result.

EDIT 1: Interestingly, the Hungarians were left out also in Tambets et al. 2018 and Peltola et al. 2023. They are within the qpAdm models by Lamnidis et al. 2018, and there they get some Nganasan ancestry - but on the other hand, so do get the English, the Scottish, and the French, too, so this does not appear to be the actual recent Siberian ancestry.

EDIT 2: For the Estonians, the Yakutian ancestry is between 3–5 % in the qpAdm models, but it can be replaced by the Mongolia or the Tyumen ancestry, but then the P-value drops. But since none of the P-values is good for the Estonians anyway, the right Siberian ancestry source for them is yet to be found. Some might suggest the Early Iron Age sample from Minino, Vologda.

[ANIEXCAVATOR]

Its very important to read the paper carefully to understand precisely what the authors meant there when they made that statement. When drawing the connection between Yakutia_LNBA and Uralic populations, the authors highlighted two sides to the relationship:

1. Yakutia_LNBA is the only East Asian ancestry found among all Uralic populations, without exception.
2. Yakutia_LNBA explains all of the East Asian ancestry found among Uralic populations.

When modeling the Uralic populations using qpAdm, the authors chose to present the simplest model that passes (i.e., the one with the smallest number of source populations) as the one closest to the truth, even if it might not capture the *entire* truth or the *most detailed* truth. This choice has to be made because there is no simple way to compare qpAdm models with different numbers of source populations. Yes, more complex models have higher p-value, but more complex models are also less parsimonious and under some contrived circumstances, at least in theory, you can keep increasing p-values by continuously increasing the number of sources.

Thank you. So, they made choices in interpretation of the data, and then they propose their own choice as the result? Sounds methodologically suspicious: "We aim for X, therefore we make choices toward X". No wonder the result is X, then.

I think the recent results concerning the Indo-Europeans show that we should not aim for the most simple explanation, because reality is rarely so economical. Unfortunately we still have no ancient DNA samples from the region and the time relevant for Proto-Uralic, so indirect approach is all we have.

Restricting the comparison to the simplest models that pass for each population, the first statement--that Yakutia_LNBA is present in all Uralic populations and is required for them to pass--is unequivocably true, and the models you posted actually show that. This continues to be the case even when you look at more complex models.

The second statement is actually *also* true, if you look at the simplest models only. If you look at more complex models, then the second statement only tends to be true.

I would say only that it could be true, but the model containing only the Yakutia ancestry is not considerably stronger than other models, based on their numeric values. That has been my point. And when the aim is to trace the language, the framework of the linguistic results could give us hints for choosing the model, instead of only aiming toward simplicity.

For example, it is known that the Proto-Samoyedic language spread from the south the north. Its homeland is located north from the Sayan Mountains, because PSy had contacts with Tocharian, Iranian, and two primary branches of Turkic. Keeping this in mind, should we not lean towards models showing also both the Mongolia and the Tyumen ancestries in the Selkups, the Enets and the Nganasans? (Quite high in the list for them all.) Interestingly, the Altai ancestry only shows in the Selkups, perhaps because they have assimilated the Yeniseian speakers.

About whether or not Srubnaya (i.e. Western Steppe European-like ancestry) is also present among all Uralic populations, the models the authors presented do not show that it is present in all Uralic populations--because Srubnaya is forced to be present in every single model, it is never an option for it to not be among the sources. Srubnaya as a generic European-like source is present in every single model for all admixed Inner Eurasian populations, to soak up the West Eurasian ancestry, because the authors were more interested in the East Asian ancestries of these Eurasian popualtions. The % of Srubnaya is very small in Nganasan and within two standard errors, indicating Nganasan has no significantly detectable western steppe ancestry.

OK, they did not compare different steppe-related ancestries, so we cannot know whether it is actually the same steppe ancestry in all the Uralic populations. Hopefully somebody sometimes studies also this topic.

Nganasans differ even from the other Samoyedic populations, being very similar to the Tundra Yukaghirs and the Dolgans (both non-Uralic populations). They were also the northernmost vanguard of the Samoyedic expansion, and the only population moving clearly beyond Yenisei to Central Siberia. Therefore it is not reasonable to consider the Nganasans as a proxy for the Proto-Samoyedic population, and much less for the Proto-Uralic population. For this same reason, the possible lack of the steppe ancestry in them is not a counter-argument against the presence of the steppe ancestry in the Proto-Uralic population.

Eh, its not circular reasoning, you've got to follow the logic as you're reading. The authors wanted to see the distribution of the two latest ancestries in the Siberian transect (Cisbaikal_LNBA and Yakutia_LNBA) among present-day Admixed Inner Eurasian populations. Because they tried everything from 2-way to 8-way models (!) they had to look at two aspects in a systematic way:

1. Out of *all passing models*, is one of the ancestries present in all of them (often one, two, three models, but oftentimes tens of models)? If so the population gets one grey dot (for Cisbaikal_LNBA) or orange dot (for Yakutia_LNBA). This shows that the ancestry is definitely present in the population.

2. For the *simplest passing model*, does either Cisbaikal_LNBA or Yakutia_LNBA alone suffice to account for all of its East Asian ancestry? If so, the population gets another grey dot or orange dot.

No population gets two grey dots, but all Yeniseian, Samoyedic and South Siberian Turkic populations get one grey dot. And Uralic-speaking populations were the only ethnolinguistic group where almost all members got two orange dots. These results emerged naturally after the procedure was decided, its not a conspiracy or something.

The p-values are not gonna exceed 0.05 in most cases because the models are extremely complex, with up to 8-way admixtures, 11 possible sources (!) and outgroup rotation (which means they had at >12 outgroups in all models already, and on top of that any model with less than 8 sources has the sources added to the outgroups). This means that the simple models (2-3 sources) have up to >20 outgroups, which is an insane number. So the authors lower the threshold to 0.01 instead (like Narasimhan, which if you recall had only up to 6-way models with 6 possible sources). That anything manages to pass at all is especially impressive because with their current setup they manage to model populations that are extremely variable from a huge area of Eurasia, while splitting the East Asian ancestries apart as well and estimating models where multiple closely-related East Asian sources are present simultaneously. This is not typically done, typically you use a few outgroups and more proximal source populations and manage to model a small number of target populations covering a limited geographic area, like in the recent PIE papers.

And the results parallel something like ADMIXTURE very closely. This kind of multi-method comparison between formal stats and other methods is also not typically done in most papers.

Did they find “Jakutia” above noise level in modern Hungarians and Estonians?
Sorry, didn’t read the article.

As you can see the model with Yakutia_LNBA is not very good for Estonians, however by removing EHG it only gets worse and the same is true if you replace EHG with Tyumen_HG.
The Estonians:
Germany_EN_LBK + Hungary_EN_HG_Koros + EHG + Yakutia_LNBA + Russia_Srubnaya = 0.03
Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02
Tyumen_HG + Germany_EN_LBK + Hungary_EN_HG_Koros + Yakutia_LNBA + Russia_Srubnaya = 0.02

If you look at all the qpAdm models, removing Yakutia_LNBA is the only thing that drops models for Uralic populations from passing to failing thresholds (>0.05 or >0.01 to <0.05 or <0.01), and furthermore removing Yakutia_LNBA is the only thing that, for all Uralic populations, produces in the order of tenfold, hundredfold or even thousandfold reductions in p-values. Switching EHG for Tyumen and so on does not produce such effects.

Did they find “Jakutia” above noise level in modern Hungarians and Estonians?
Sorry, didn’t read the article.

For some reason they did not include the Hungarians in their calculations. One might suspect that was because the modern Hungarians would not show any traces of the Yakutia ancestry, thus nullifying their pre-decided result.
EDIT 1: Interestingly, the Hungarians were left out also in Tambets et al. 2018 and Peltola et al. 2023. They are within the qpAdm models by Lamnidis et al. 2018, and there they get some Nganasan ancestry - but on the other hand, so do get the English, the Scottish, and the French, too, so this does not appear to be the actual recent Siberian ancestry.
EDIT 2: For the Estonians, the Yakutian ancestry is between 3–5 % in the qpAdm models, but it can be replaced by the Mongolia or the Tyumen ancestry, but then the P-value drops. But since none of the P-values is good for the Estonians anyway, the right Siberian ancestry source for them is yet to be found. Some might suggest the Early Iron Age sample from Minino, Vologda.

The authors were looking for Admixed *Inner Eurasian* populations, populations that are part of the clines stretching from Europeans/West Asians to East Asians/Siberians, containing most Central Asian and North Eurasian populations. Hungarians are a generic Central European population and are not a part of these clines, and there are other populations that the authors exclude too (e.g. Turks from Anatolia). This is once again imputing motives to the authors that on an unbiased close reading they simply don't have.

And this leaves aside the fact that we already know that the arrival of Magyars into the Pannonian Plain involved the sudden increase in levels of Yakutia_LNBA ancestry and haplogroup N compared to those previously there (the Avars), so this isn't an exception to the trend the authors have stated.

[ANIEXCAVATOR]

About the West Eurasian ancestries among Uralic populations, we already know they are gonna be very heterogeneous because, for example in this paper, we already see that WHG ancestry is present towards the west (i.e., West Uralic populations have Slavic- or Baltic-like West Eurasian ancestry) that is not present elsewhere, and Nganasan have almost no West Eurasian ancestry/no West Eurasian ancestry at all so even CW-related, Fatyanovo or Abashevo-related ancestry can only be present only among all Uralics except Nganasan, if it is indeed present among so many of them at all. So a "universal West Eurasian ancestry" among Uralic populations is already not present, a moot point. This is already apparent from loads of ADMIXTURE and formal stat-type analyses from many papers and even in amateur analyses. It would of course still be a good idea for a peer-review, published paper to explictly state this to get this out into the record.

[Queequeg]

...And this leaves aside the fact that we already know that the arrival of Magyars into the Pannonian Plain involved the sudden increase in levels of Yakutia_LNBA ancestry and haplogroup N compared to those previously there (the Avars), so this isn't an exception to the trend the authors have stated.

Many thanks for your important comments, ANIEXCAVATOR. Now that you mention a sudden increase, do we know that Avars were, at least to some extent, based on the same Yakutia_LNBA?

[Jaska]

No population gets two grey dots, but all Yeniseian, Samoyedic and South Siberian Turkic populations get one grey dot. And Uralic-speaking populations were the only ethnolinguistic group where almost all members got two orange dots. These results emerged naturally after the procedure was decided, its not a conspiracy or something.

Naturally I do not assume any conspiracy here. I just say that their results do not seem to justify the exclusion of all the other eastern ancestries and the picking up of only the simplest model where there is no other eastern ancestries but only the Yakutia ancestry in the Uralic populations. As we can see, in many populations there could just as well be several eastern ancestries present. This ambiguity of the results should not be just ignored in order to aim for the simplest possible model.

It looks plausible that the Yakutia ancestry is needed in all the Uralic populations (and in some others, too), but this does not automatically lead to the conclusion that it was the sole eastern ancestry.

Thanks for explaining their process thoroughly.

The authors were looking for Admixed *Inner Eurasian* populations, populations that are part of the clines stretching from Europeans/West Asians to East Asians/Siberians, containing most Central Asian and North Eurasian populations. Hungarians are a generic Central European population and are not a part of these clines, and there are other populations that the authors exclude too (e.g. Turks from Anatolia). This is once again imputing motives to the authors that on an unbiased close reading they simply don't have.

The Hungarians are a Uralic speaking population, exclusion of which looks weird when they make claims concerning especially the Uralic populations. You must admit that if they had the Hungarians included, they probably had to rephrase their main conclusion about the Yakutia ancestry in the modern Uralic populations (although we see the eastern ancestries in the ancient Hungarians).

And how “inner Eurasian” is the Kola Peninsula? Still they included the BOO samples, probably because there was a possible connection to the Uralic speaking populations. Yet they excluded the Hungarians, who are undoubtedly Uralic speaking.

[Jaska]

About the West Eurasian ancestries among Uralic populations, we already know they are gonna be very heterogeneous because, for example in this paper, we already see that WHG ancestry is present towards the west (i.e., West Uralic populations have Slavic- or Baltic-like West Eurasian ancestry) that is not present elsewhere, and Nganasan have almost no West Eurasian ancestry/no West Eurasian ancestry at all so even CW-related, Fatyanovo or Abashevo-related ancestry can only be present only among all Uralics except Nganasan, if it is indeed present among so many of them at all. So a "universal West Eurasian ancestry" among Uralic populations is already not present, a moot point. This is already apparent from loads of ADMIXTURE and formal stat-type analyses from many papers and even in amateur analyses. It would of course still be a good idea for a peer-review, published paper to explictly state this to get this out into the record.

About the "universal West Eurasian ancestry among Uralic populations is already not present": 
Of course we do not assume the WHG ancestry in all the Uralic populations, but there were already EHG and the steppe ancestry in the Kama-Ural Region during the Proto-Uralic time - widespread in the Uralic populations in both the qpAdm and the Admixture results. 

I already wrote about how aberrant population the Nganasans are even among the Samoyeds, so they are not a counter-argument concerning the presence of ancestries in the Proto-Uralic population - any more than the lack of the Yakutia ancestry in the modern Hungarians is a counter-argument against the presence of this ancestry in the Proto-Uralic population.

[Anglesqueville]

The right List is Zeng's but reduced to what seems to me to be not completely useless (although we could reduce it further). As always I use Admixtools (and not Admixtools 2), qpfstats, allSNPs = YES. I kept the transitions because otherwise, the number of Nganasan SNPs would become insufficient. You will notice the lack of a model for the Finns. The reason is that all models lead to ridiculous tail probs, and I suspect an error for them when I built the files. But I believe I have posted models elsewhere for the Finns and the Saamis which are much better than these distant models (effectively based on Minino_IA). Regarding your discussion, the absence of an Indo-European source in the Nganassans seems to me symmetrical with the absence of a Siberian source in the Hungarians. In either case, genetic dilution contrasts with linguistic transmission. Personally, I don't see anything shocking in it.


right pops:
Siberia_UpperPaleolithic_UstIshim
Siberia_UpperPaleolithic_Malta
Russia_UpperPaleolithic_Kostenki
Siberia_UpperPaleolithic_Yana
Iran_Neolithic
Anatolia_Neolithic_Aceramic
China_AmurRiver_LPaleolithic
China_Tianyuan
Italy_South_HG_Ostuni1
China_NEastAsia_Coastal_EN
China_SEastAsia_Island_EN
USA_AK_NeoAleut
USA_AK_PaleoAleut.SG


1) Hungarians

left pops:
Hungarian.DG
Russia_BronzeAge_Srubnaya
LBK_Halberstadt
Hungary_Neolithic_Koros
Yakutia_LN

best coefficients:    0.590    0.331    0.073    0.006
totmean:      0.590    0.331    0.073    0.006
boot mean:    0.590    0.331    0.072    0.006
      std. errors:    0.058    0.040    0.052    0.010


fixed pat  wt  dof    chisq      tail prob
        0000  0    9    5.773        0.762358    0.590    0.331    0.073    0.006
        0001  1    10    6.210        0.797349    0.606    0.326    0.068    0.000
        0010  1    10    7.968        0.63201    0.646    0.349    0.000    0.004
       
best pat:        0000        0.762358              -  -
best pat:        0001        0.797349  chi(nested):    0.436 p-value for nested model:        0.508956
best pat:        0011        0.692378  chi(nested):    2.022 p-value for nested model:        0.154998


left pops:
Hungarian.DG
Russia_BronzeAge_Srubnaya
LBK_Halberstadt
Hungary_Neolithic_Koros


best coefficients:    0.603    0.327    0.070
totmean:      0.603    0.327    0.070
boot mean:    0.604    0.327    0.070
      std. errors:    0.051    0.039    0.050

fixed pat  wt  dof    chisq      tail prob
          000  0    10    5.692        0.840407    0.603    0.327    0.070
          001  1    11    7.680        0.741691    0.654    0.346    0.000
     
     
best pat:          000        0.840407              -  -
best pat:          001        0.741691  chi(nested):    1.987 p-value for nested model:        0.158646

2) Nganasans

left pops:
Nganasan
Tyumen_HG
Mongolia_N_North
Yakutia_LN
Russia_BronzeAge_Srubnaya


best coefficients:    -0.114    0.488    0.644    -0.019
totmean:    -0.114    0.488    0.644    -0.019
boot mean:    -0.580    0.604    1.036    -0.060
      std. errors:    23.910    7.617    18.776    2.322


fixed pat  wt  dof    chisq      tail prob
        0000  0    9    10.473        0.313582    -0.114    0.488    0.644    -0.019  infeasible
        0001  1    10    12.239        0.26938    -0.006    0.391    0.615    0.000  infeasible
        0010  1    10    36.029    8.32014e-05    0.539    0.426    0.000    0.035
        0100  1    10    17.443      0.0651105    -0.009    0.000    0.986    0.023  infeasible
        1000  1    10    12.410        0.258558    0.000    0.421    0.585    -0.006  infeasible
        0011  2    11    36.818    0.000123595    0.345    0.655    0.000    0.000
        0101  2    11    23.068      0.0172899    -0.256    0.000    1.256    0.000  infeasible
        0110  2    11    38.844    5.63106e-05    0.907    0.000    0.000    0.093
        1001  2    11    15.337        0.167583    0.000    0.405    0.595    0.000
        1010  2    11    42.801    1.17638e-05    0.000    1.050    0.000    -0.050  infeasible
        1100  2    11    19.823      0.0478224    0.000    0.000    0.975    0.025
        0111  3    12    84.541    5.57642e-13    1.000    0.000    0.000    0.000
        1011  3    12    66.484    1.44797e-09    0.000    1.000    0.000    0.000
        1101  3    12    23.459      0.0240731    0.000    0.000    1.000    0.000
        1110  3    12  5205.604              0    0.000    0.000    0.000    1.000
best pat:        0000        0.313582              -  -
best pat:        1000                0  chi(nested):    1.937 p-value for nested model:        0.163981


left pops:
Nganasan
Mongolia_N_North
Yakutia_LN
Russia_BronzeAge_Srubnaya


best coefficients:    0.416    0.590    -0.006
totmean:      0.416    0.590    -0.006
boot mean:    0.412    0.594    -0.006
      std. errors:    0.168    0.157    0.015
     
fixed pat  wt  dof    chisq      tail prob
          000  0    10    10.032        0.437726    0.416    0.590    -0.006  infeasible
          001  1    11    10.517        0.484547    0.380    0.620    0.000
         
         
left pops:
Nganasan
Mongolia_N_North
Yakutia_LN


best coefficients:    0.378    0.622
totmean:      0.378    0.622
boot mean:    0.370    0.630
      std. errors:    0.102    0.102
     
     
fixed pat  wt  dof    chisq      tail prob
          00  0    11    9.624        0.564526    0.378    0.622
          01  1    12    64.896    2.84922e-09    1.000    0.000
          10  1    12    19.860      0.0697812    0.000    1.000
best pat:          00        0.564526              -  -
best pat:          10        0.0697812  chi(nested):    10.236 p-value for nested model:      0.00137698


3) Saamis

left pops:
Utsjoki_Saami
Russia_BronzeAge_Srubnaya
LBK_Halberstadt
Yakutia_LN
Russia_Mesolithic >>>>> Sidelkino

best coefficients:    0.331    0.170    0.256    0.243
totmean:      0.331    0.170    0.256    0.243
boot mean:    0.330    0.171    0.256    0.244
      std. errors:    0.176    0.091    0.014    0.088


fixed pat  wt  dof    chisq      tail prob
        0000  0    9    9.219        0.417264    0.331    0.170    0.256    0.243
        0001  1    10    20.145      0.0279097    0.792    -0.045    0.253    0.000  infeasible
        0010  1    10    79.561    6.12141e-13    14.945    -7.177    0.000    -6.767  infeasible
        0100  1    10    13.889        0.178133    0.641    0.000    0.251    0.108
        1000  1    10    13.897        0.177739    0.000    0.336    0.262    0.402


4) Estonians

left pops:
Estonians
Russia_BronzeAge_Srubnaya
LBK_Halberstadt
Hungary_Neolithic_Koros
Yakutia_LN


best coefficients:    0.766    0.151    0.072    0.011
totmean:      0.766    0.151    0.072    0.011
boot mean:    0.767    0.151    0.071    0.011
      std. errors:    0.058    0.041    0.050    0.010


fixed pat  wt  dof    chisq      tail prob
        0000  0    9    6.712        0.667045    0.766    0.151    0.072    0.011
        0001  1    10  202.582              0    1.564    0.392    -0.956    0.000  infeasible
        0010  1    10    8.925        0.539228    0.820    0.169    0.000    0.011
        0100  1    10    19.713      0.0320862    0.868    0.000    0.128    0.004
        1000  1    10  128.733    8.46797e-23    0.000    0.284    0.661    0.055
        0011  2    11    10.205        0.512071    0.839    0.161    0.000    0.000
       
       
left pops:
Estonians
Russia_BronzeAge_Srubnaya
LBK_Halberstadt
Hungary_Neolithic_Koros


best coefficients:    0.787    0.144    0.069
totmean:      0.787    0.144    0.069
boot mean:    0.788    0.143    0.068
      std. errors:    0.054    0.041    0.051


fixed pat  wt  dof    chisq      tail prob
          000  0    10    7.591        0.668699    0.787    0.144    0.069
          001  1    11    9.557        0.570651    0.839    0.161    0.000