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

[Jaska]

Wow, thank you very much, Anglesqueville!
So, the Hungarians work best without any Yakutia ancestry, and the Nganasans without any steppe ancestry.

Earlier studies have shown for the Saami ca. 50 % of the Corded Ware mediated ancestry, but here it seems that half of that could rather be direct EHG ancestry? But then, considering your models for the Levänluhta population, how is the Minino ancestry divided between the CWC and the direct EHG ancestries?

[Kale]

That's not a terribly tight suite of right pops. There is no WHG for example, and if you check F4 stats, WHG are more related to Estonians than to Srubnaya, and wayyyy more related to Estonians than either of the farmers. E.g. adding a WHG to the right would cause the model to fail. Unless the Koros there is the HG outlier?

[Anglesqueville]

^^Kale, I wanted to stick as closely as possible to Zeng, so in particular I took his right list trying to give it a less extravagant size, but without adding anything (except, I seem to remember having added one or two Paleolithic sources, perhaps useless). I agree that this right list is unbalanced, in particular very poor in WHG. As for the choice of the Sicilian WHG, I admit that I am perplexed.
Jaska, I will think about your question, but completely independently of Zeng.

[Jaska]

I checked the composition of the Early Iron Age Minino NEO538 in Allentoft et al. 2024:

In supp. table IX, Deep:
MiddleDon_7500BP = 0,653
Caucasus_13000BP_10000BP = 0,072
Boncuklu_10000BP = 0,015
SiberiaNE_9800BP = 0,144
Amur_7500BP = 0,116

In supp. table X, PostNeol:
MiddleDon_7500BP = 0,387
Poland_5000BP_4700BP = 0,027
Steppe_5000BP_4300BP = 0,305
SteppeCE_7000BP_3600BP = 0,06
SiberiaNE_9800BP = 0,069
Amur_7500BP = 0,083
Baikal_8000BP_7200BP = 0,069

In Supp. table XI, PostBA:
Scandinavia_4000BP_3000BP = 0,049
Estonia_3000BP_2500BP = 0,262
EurasiaN_2000BP_1000BP = 0,647
Amur_7500BP = 0,042

[Anglesqueville]

 But then, considering your models for the Levänluhta population, how is the Minino ancestry divided between the CWC and the direct EHG ancestries?

right pops:
Russia_Ust_Ishim.DG
Italy_North_Villabruna_HG
Czech_Vestonice16
Belgium_UP_GoyetQ116_1
Russia_MA1_HG.SG
Iran_GanjDareh_N
Russia_Kostenki14.SG
Indian_GreatAndaman_100BP.SG
Israel_PPNB
Yana.ial
Georgia_mesolithic
Amur_neo.ial
Kolyma_meso.ial
Turkey_N_I0707

left pops:
Russia_Minino_IronAge.ial
Estonia_BA_ial
Sidelkino.ial
Yakutia_LN

best coefficients:    0.360    0.211    0.430
totmean:      0.360    0.211    0.430
boot mean:    0.359    0.211    0.430
      std. errors:    0.042    0.044    0.018
     
fixed pat  wt  dof    chisq      tail prob
          000  0    12    11.240        0.508486    0.360    0.211    0.430
          001  1    13  413.663              0    -1.721    2.721    0.000  infeasible
          010  1    13    34.579    0.000982236    0.538    0.000    0.462
          100  1    13    72.792    2.44781e-10    0.000    0.568    0.432
          011  2    14  779.318              0    1.000    0.000    0.000
          101  2    14  480.321              0    0.000    1.000    0.000
          110  2    14  767.470              0    0.000    0.000    1.000
best pat:          000        0.508486              -  -
best pat:          010      0.000982236  chi(nested):    23.339 p-value for nested model:    1.35797e-06
best pat:          101      1.36852e-93 not nested

This model is close to Allentoft's third (post_BA), at least in terms of its composition. When two methods as different as Admixture and qpAdm converge like this I think we can count on something solid. And then 0.5 is a really good tail prob and the standard errors are as tight as possible. I believe the answer is in this model, at least in substance. I tried other choices than Sidelkino without finding anything really better.

[kolompar]

Alright, so Supplementary 7 Table 1 is the right list? Really more focused on the Eastern ancestry and Srubnaya is just a stand-in for something Steppe-related and shouldn't be taken seriously.
But for Angles there are quite a few European HGs in there that should be relevant, at least more so than Ostuni...
And there should be some punishment for still bringing up Hungarians as an example in 2024 It's always been clear what happened there, and that they are the exception to the rule, and now we have concrete genetic evidence too. I don't see how the case of a few hundred Nganasans scattered in the tundra would be anything like the Hungarians conquering a large population and establishing their state.
It's not even true that Hungarians don't have detectable Siberian ancestry, I posted models for it before https://genarchivist.com/showthread.php?...09#pid8409
Even a simple admixture f3 will show it too:

Code:

  pop1         pop2              pop3            est      se     z      p      n
  <chr>        <chr>             <chr>         <dbl>   <dbl> <dbl>  <dbl>  <dbl>
1 Hungarian.HO Nganasan.HO       Czech.HO  -0.000773 3.85e-4 -2.01 0.0443 593124
2 Hungarian.HO Nganasan.HO       Croatian~ -0.000738 3.93e-4 -1.88 0.0608 593124
3 Hungarian.HO Russia_Krasnoyar~ Czech.HO  -0.00102  5.67e-4 -1.80 0.0719 592866
4 Hungarian.HO Russia_Krasnoyar~ Croatian~ -0.000888 6.01e-4 -1.48 0.140  592866
5 Hungarian.HO Russia_KolymaRiv~ Czech.HO  -0.000875 4.83e-4 -1.81 0.0704 592999
6 Hungarian.HO Russia_KolymaRiv~ Croatian~ -0.000764 4.93e-4 -1.55 0.122  592999

And qpAdm in the same vein, right = c("Russia_Ust_Ishim.DG", "Turkey_N", "Serbia_IronGates_Mesolithic", "Russia_Samara_EBA_Yamnaya", "Russia_Shamanka_Eneolithic.SG", "Russia_Arkhangelsk_Veretye_Mesolithic.SG", "Russia_Kolyma_M.SG", "Russia_LenaRiver_EN.SG", "Russia_Karelia_HG", "Kazakhstan_Botai_Eneolithic.SG", "Russia_Tyumen_HG", "USA_Ancient_Beringian.SG", "Russia_AngaraRiver_EN.SG", "Russia_DevilsCave_N.SG", "Mongolia_North_N", "China_YR_MN", "Russia_Yaroslavl_Fatyanovo_BA.SG", "Russia_Mezhovskaya.SG", "Russia_Srubnaya", "Russia_CentralYakutia_LN.SG").

Code:

$weights
# A tibble: 2 x 5
  target       left                     weight     se     z
  <chr>        <chr>                     <dbl>  <dbl> <dbl>
1 Hungarian.DG Czech.DG                 0.986  0.0101 97.2
2 Hungarian.DG Russia_Krasnoyarsk_BA.SG 0.0141 0.0101  1.39

$rankdrop
# A tibble: 2 x 7
  f4rank   dof  chisq     p dofdiff chisqdiff p_nested
   <int> <int>  <dbl> <dbl>   <int>     <dbl>    <dbl>
1      1    18   19.0 0.393      20     5977.        0
2      0    38 5996.  0          NA       NA        NA

$popdrop
# A tibble: 3 x 13
  pat      wt   dof  chisq     p f4rank Czech.DG Russia_Krasnoyarsk_B~ feasible
  <chr> <dbl> <dbl>  <dbl> <dbl>  <dbl>    <dbl>                 <dbl> <lgl>   
1 00        0    18   19.0 0.393      1    0.986                0.0141 TRUE   
2 01        1    19   22.8 0.246      0    1                   NA      TRUE   
3 10        1    19 5882.  0          0   NA                    1      TRUE   
# ... with 4 more variables: best <lgl>, dofdiff <dbl>, chisqdiff <dbl>,
#   p_nested <dbl>

Code:

$weights
# A tibble: 3 x 5
  target       left                      weight      se     z
  <chr>        <chr>                      <dbl>   <dbl> <dbl>
1 Hungarian.HO Czech.HO                 0.513   0.108    4.76
2 Hungarian.HO Croatian.HO              0.477   0.108    4.41
3 Hungarian.HO Russia_Krasnoyarsk_BA.SG 0.00955 0.00289  3.30

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq        p dofdiff chisqdiff    p_nested
   <int> <int>  <dbl>    <dbl>   <int>     <dbl>       <dbl>
1      2    17   15.1 0.587         19      60.1  0.00000376
2      1    36   75.2 0.000140      21    6376.   0        
3      0    57 6451.  0             NA      NA   NA        

$popdrop
# A tibble: 7 x 14
  pat      wt   dof  chisq        p f4rank Czech.HO Croatian.HO Russia_Krasnoya~
  <chr> <dbl> <dbl>  <dbl>    <dbl>  <dbl>    <dbl>       <dbl>            <dbl>
1 000       0    17   15.1 0.587         2    0.513       0.477          0.00955
2 001       1    18   23.3 0.178         1    0.491       0.509         NA     
3 010       1    18   31.4 0.0260        1    0.988      NA              0.0125
4 100       1    18   37.6 0.00441       1   NA           0.993          0.00747
5 011       2    19   47.0 0.000363      0    1          NA             NA     
6 101       2    19   42.5 0.00150       0   NA           1             NA     
7 110       2    19 6309.  0             0   NA          NA              1     
# ... with 5 more variables: feasible <lgl>, best <lgl>, dofdiff <dbl>,
#   chisqdiff <dbl>, p_nested <dbl>

Code:

$weights
# A tibble: 2 x 5
  target     left                     weight     se     z
  <chr>      <chr>                     <dbl>  <dbl> <dbl>
1 Finnish.DG Norwegian.DG             0.953  0.0102 93.8
2 Finnish.DG Russia_Krasnoyarsk_BA.SG 0.0467 0.0102  4.60

$rankdrop
# A tibble: 2 x 7
  f4rank   dof  chisq     p dofdiff chisqdiff p_nested
   <int> <int>  <dbl> <dbl>   <int>     <dbl>    <dbl>
1      1    18   20.2 0.323      20     5505.        0
2      0    38 5526.  0          NA       NA        NA

$popdrop
# A tibble: 3 x 13
  pat      wt   dof  chisq       p f4rank Norwegian.DG Russia_Krasnoyarsk_BA.SG
  <chr> <dbl> <dbl>  <dbl>   <dbl>  <dbl>        <dbl>                    <dbl>
1 00        0    18   20.2 0.323        1        0.953                   0.0467
2 01        1    19   43.6 0.00108      0        1                      NA     
3 10        1    19 5315.  0            0       NA                       1     
# ... with 5 more variables: feasible <lgl>, best <lgl>, dofdiff <dbl>,
#   chisqdiff <dbl>, p_nested <dbl>

Code:

$weights
# A tibble: 3 x 5
  target     left                     weight      se      z
  <chr>      <chr>                     <dbl>   <dbl>  <dbl>
1 Finnish.HO Norwegian.HO             0.0187 0.116    0.161
2 Finnish.HO Estonian.HO              0.925  0.118    7.86
3 Finnish.HO Russia_Krasnoyarsk_BA.SG 0.0559 0.00558 10.0 

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq        p dofdiff chisqdiff  p_nested
   <int> <int>  <dbl>    <dbl>   <int>     <dbl>     <dbl>
1      2    17   21.3 2.13e- 1      19      112.  3.10e-15
2      1    36  134.  3.93e-13      21     6044.  0       
3      0    57 6178.  0             NA       NA  NA       

$popdrop
# A tibble: 7 x 14
  pat      wt   dof  chisq        p f4rank Norwegian.HO Estonian.HO
  <chr> <dbl> <dbl>  <dbl>    <dbl>  <dbl>        <dbl>       <dbl>
1 000       0    17   21.3 2.13e- 1      2       0.0187       0.925
2 001       1    18   71.0 3.08e- 8      1      -1.53         2.53
3 010       1    18  116.  2.84e-16      1       0.928       NA   
4 100       1    18   27.3 7.37e- 2      1      NA            0.943
5 011       2    19  490.  6.05e-92      0       1           NA   
6 101       2    19  249.  3.96e-42      0      NA            1   
7 110       2    19 5456.  0             0      NA           NA   
# ... with 6 more variables: Russia_Krasnoyarsk_BA.SG <dbl>, feasible <lgl>,
#   best <lgl>, dofdiff <dbl>, chisqdiff <dbl>, p_nested <dbl>

Code:

$weights
# A tibble: 2 x 5
  target      left                     weight      se      z
  <chr>       <chr>                     <dbl>   <dbl>  <dbl>
1 Estonian.HO Lithuanian.HO            0.980  0.00416 236. 
2 Estonian.HO Russia_Krasnoyarsk_BA.SG 0.0202 0.00416   4.87

$rankdrop
# A tibble: 2 x 7
  f4rank   dof  chisq     p dofdiff chisqdiff p_nested
   <int> <int>  <dbl> <dbl>   <int>     <dbl>    <dbl>
1      1    18   21.4 0.262      20     5958.        0
2      0    38 5979.  0          NA       NA        NA

$popdrop
# A tibble: 3 x 13
  pat      wt   dof  chisq        p f4rank Lithuanian.HO Russia_Krasnoyarsk_BA~
  <chr> <dbl> <dbl>  <dbl>    <dbl>  <dbl>         <dbl>                  <dbl>
1 00        0    18   21.4 0.262         1         0.980                 0.0202
2 01        1    19   45.2 0.000637      0         1                    NA     
3 10        1    19 5816.  0             0        NA                     1     
# ... with 5 more variables: feasible <lgl>, best <lgl>, dofdiff <dbl>,
#   chisqdiff <dbl>, p_nested <dbl>

[Jaska]

It's always been clear what happened there, and that they are the exception to the rule, and now we have concrete genetic evidence too. I don't see how the case of a few hundred Nganasans scattered in the tundra would be anything like the Hungarians conquering a large population and establishing their state.

Why should it have been similar as a process? Nobody claimed that. It can be similar as a result: different processes can lead to similar results. The fact remains that the Nganasans genetically greatly resemble the (non-Uralic) Tundra Yukaghirs and Dolgans, clearly differing from the other Samoyedic populations.

[Queequeg]

It can be similar as a result: different processes can lead to similar results.

This is a good point and it may well be that Nganasans have been sort of developing back towards pure Yakutia_LNBA type of a mixture through a later, post Proto Uralic level admixture. On the Proto Uralic level something like Yakutia_LNBA and EHG-ish, maybe still leaning towards Steppe groups through contacts with Indo Iranians, looks more appropriate. I however suspect that the original Siberian root of the Proto Uralic speaking group was somewhat more southern than the current Yakutia_LNBA group, having less Yakutia_MN, but that remains to be seen. Apparently the Tatarka Hill samples and ROT002 of Rostovka have not yet been published?

Then, it seems to me that after the Proto Uralic level, during and after the dissolution of the Koptyaki complex, as Jaska has sort of suggested, the western groups slowly develop into Ananyino Culture (more Baltic_BA, looking after that like NEO538) on the European side of Ural mountains and into Sargat Culture (more IA_Steppe, looking then like Sargat_SMV001, i.e. first not even very far from NEO538) on the eastern side of the mountains, the related sub cultures of course included.

[kolompar]

It's always been clear what happened there, and that they are the exception to the rule, and now we have concrete genetic evidence too. I don't see how the case of a few hundred Nganasans scattered in the tundra would be anything like the Hungarians conquering a large population and establishing their state.

Why should it have been similar as a process? Nobody claimed that. It can be similar as a result: different processes can lead to similar results. The fact remains that the Nganasans genetically greatly resemble the (non-Uralic) Tundra Yukaghirs and Dolgans, clearly differing from the other Samoyedic populations.

I replied to Angles making the comparison. Hungarians are a very rare case, possibly the only one, having their language's ancestry diluted so much. I don't know why we would assume a different process with no precedent and no evidence. But right, you would need something even more special to spread language over a million square kilometers with no geneflow.
I'm sure even you don't believe what you say, Evenk, Dolgan, Yakut are known to be recent developments and have obvious extra ancestry that brought their language. Yukaghir aren't all that similar either, I couldn't model Nganasan with them but go ahead and show us if you could.
Samoyeds on the other hand are quite straightforward:

Code:

$weights
# A tibble: 3 x 5
  target   left                     weight     se     z
  <chr>    <chr>                     <dbl>  <dbl> <dbl>
1 Enets.HO Russia_Krasnoyarsk_BA.SG  0.490 0.0525  9.33
2 Enets.HO Ket.HO                    0.398 0.0763  5.22
3 Enets.HO Russian.HO                0.112 0.0265  4.24

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq        p dofdiff chisqdiff  p_nested
   <int> <int>  <dbl>    <dbl>   <int>     <dbl>     <dbl>
1      2    17   21.2 2.18e- 1      19      292.  7.14e-51
2      1    36  314.  5.26e-46      21     9135.  0      
3      0    57 9449.  0             NA       NA  NA      

Code:

$weights
# A tibble: 3 x 5
  target    left                     weight     se     z
  <chr>     <chr>                     <dbl>  <dbl> <dbl>
1 Selkup.HO Russia_Krasnoyarsk_BA.SG 0.0785 0.0244  3.21
2 Selkup.HO Ket.HO                   0.566  0.0686  8.26
3 Selkup.HO Mansi.HO                 0.355  0.0467  7.61

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq        p dofdiff chisqdiff  p_nested
   <int> <int>  <dbl>    <dbl>   <int>     <dbl>     <dbl>
1      2    17   15.4 5.64e- 1      19      121.  7.36e-17
2      1    36  136.  1.34e-13      21     2127.  0      
3      0    57 2263.  0             NA       NA  NA      

It can be similar as a result: different processes can lead to similar results.

This is a good point  and it may well be that Nganasans have been sort of developing back towards pure Yakutia_LNBA type of a mixture through a later, post Proto Uralic level admixture. On the Proto Uralic level something like Yakutia_LNBA and EHG-ish, maybe still leaning towards Steppe groups through contacts with Indo Iranians, looks more appropriate. I however suspect that the original Siberian root of the Proto Uralic speaking group was somewhat more southern than the current Yakutia_LNBA group, having less Yakutia_MN, but that remains to be seen. Apparently the Tatarka Hill samples and ROT002 of Rostovka have not yet been published?

Then, it seems to me that after the Proto Uralic level, during and after the dissolution of the Koptyaki complex, as Jaska has sort of suggested, the western groups slowly develop into Ananyino Culture (more Baltic_BA, looking after that like NEO538) on the European side of Ural mountains and into Sargat Culture (more IA_Steppe, looking then like Sargat_SMV001, i.e. first not even very far from NEO538) on the eastern side of the mountains, the related sub cultures of course included.

I don't doubt Nganasans could be partly earlier Uralo-Yukaghir, but is there any evidence of other ancestry in Proto-Uralic? I've posted models for several populations where kra001 alone is sufficient. And I agree that Childebayeva preprint is quite interesting, a fully Siberian individual, after 2000BC, far west from the similar samples, close to modern Uralic territory, next to all those possible Indo-Iranians... What is a Yukaghir doing there?

[Queequeg]

I don't doubt Nganasans could be partly earlier Uralo-Yukaghir, but is there any evidence of other ancestry in Proto-Uralic? I've posted models for several populations where kra001 alone is sufficient. And I agree that Childebayeva preprint is quite interesting, a fully Siberian individual, after 2000BC, far west from the similar samples, close to modern Uralic territory, next to all those possible Indo-Iranians... What is a Yukaghir doing there?

I'm sorry, I don't quite understand what you're trying to say. Or, are you basically just saying that Proto Uralic was spoken by pure Yakutia_LNBA or some part of that? I'm not BTW aware of any proven deeper connections between Uralic and Yukaghir.

[Jaska]

I replied to Angles making the comparison. Hungarians are a very rare case, possibly the only one, having their language's ancestry diluted so much. I don't know why we would assume a different process with no precedent and no evidence.

What do you mean by this? I meant that the process concerning the Nganasans certainly was different from the Hungarians, just like you said.

But right, you would need something even more special to spread language over a million square kilometers with no geneflow.

You made a strawman. Nobody has proposed a language spread with no geneflow.

I'm sure even you don't believe what you say, Evenk, Dolgan, Yakut are known to be recent developments and have obvious extra ancestry that brought their language. Yukaghir aren't all that similar either, I couldn't model Nganasan with them but go ahead and show us if you could.

Look at the qpAdm results by Zeng et al. 2023: you will see several ancestries shared by the Nganasans and those populations.

Samoyeds on the other hand are quite straightforward:

You are using modern populations as the sources.

I don't doubt Nganasans could be partly earlier Uralo-Yukaghir, but is there any evidence of other ancestry in Proto-Uralic? I've posted models for several populations where kra001 alone is sufficient. And I agree that Childebayeva preprint is quite interesting, a fully Siberian individual, after 2000BC, far west from the similar samples, close to modern Uralic territory, next to all those possible Indo-Iranians... What is a Yukaghir doing there?

Could you tell us your method: how can you see the language from the DNA? So far no scientist has been able to invent such a method, because language actually is not inherited in the DNA.

[kolompar]

Gave admixtools2's qpadm_rotate a try, right pops were:

Code:

Russia_Ust_Ishim.DG
Turkey_N
Serbia_IronGates_Mesolithic
Russia_Samara_EBA_Yamnaya
China_NEastAsia_Coastal_EN
Russia_Arkhangelsk_Veretye_Mesolithic.SG
Russia_Kolyma_M.SG
Russia_LenaRiver_N.SG
Russia_Shamanka_Eneolithic.SG
USA_Ancient_Beringian.SG
Russia_AngaraRiver_EN.SG
Russia_DevilsCave_N.SG
China_YR_MN
Iran_GanjDareh_N

Russia_Karelia_HG
Kazakhstan_Botai_Eneolithic.SG
Russia_Tyumen_HG
Mongolia_North_N
Russia_Yaroslavl_Fatyanovo_BA.SG
Russia_Mezhovskaya.SG
Russia_Srubnaya
Ukraine_Mesolithic
Turkmenistan_Gonur_BA_1
Russia_AngaraRiver_BA.SG
China_YR_LBIA
China_AmurRiver_N
Russia_CentralYakutia_LN.SG

With the first bunch being fixed and the second one rotated. This was already 8000+ models and took hours to run.

For Nganasan the best feasible model was Russia_AngaraRiver_BA.SG + China_YR_LBIA + Russia_CentralYakutia_LN.SG with p = 0.482 but it was also a pass without YR_LBIA (p=0.377) and with just Yakutia too (p=0.24).
For Enets the best was Russia_Karelia_HG + Russia_Mezhovskaya.SG + Russia_AngaraRiver_BA.SG + Russia_CentralYakutia_LN.SG but p value was already only 0.008, next best models were similar with the same 4 pops, either without Karelia or Angara or with YR_LBIA or Mongolia_North_N added.
I tried Mansi next, even the best non feasible p was only 0.046, best feasible model is Russia_Karelia_HG + Russia_Mezhovskaya.SG + Russia_CentralYakutia_LN.SG p=0.000125.
So the pattern is clearly there but the models aren't exactly successful. The problem is probably the lack of good European source, maybe also too much overlap in EHG/WSHG. I won't really try more as every extra pop doubles the number of models and that takes too much time to run.

I will prefer my models with moderns, they are sensible and successful with all the above populations in the right.
Best I found for Nganasan:

Code:

$weights
# A tibble: 2 x 5
  target      left                     weight     se     z
  <chr>       <chr>                     <dbl>  <dbl> <dbl>
1 Nganasan.HO Russia_Krasnoyarsk_BA.SG  0.671 0.0493 13.6
2 Nganasan.HO Todzin.HO                 0.329 0.0493  6.69

$rankdrop
# A tibble: 2 x 7
  f4rank   dof chisq        p dofdiff chisqdiff  p_nested
   <int> <int> <dbl>    <dbl>   <int>     <dbl>     <dbl>
1      1    25  27.2 3.48e- 1      27      363.  1.98e-60
2      0    52 390.  3.05e-53      NA       NA  NA      

Actually makes the most sense that they would have Turkic ancestry, the half million Yakuts nearby didn't come from nothing. Todzin are from Tuva, should be a good proxy for the original proto/Siberian Turkic speakers. Maybe even N-P43 comes from there?

Other Samoyeds as above, kra001 base, known Yeniseian admixture, plus various more Western sources work.

Code:

$weights
# A tibble: 3 x 5
  target   left                     weight     se     z
  <chr>    <chr>                     <dbl>  <dbl> <dbl>
1 Enets.HO Russia_Krasnoyarsk_BA.SG  0.394 0.0311 12.7
2 Enets.HO Ket.HO                    0.398 0.0603  6.60
3 Enets.HO Yukagir_Forest.HO         0.208 0.0372  5.59

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq        p dofdiff chisqdiff  p_nested
   <int> <int>  <dbl>    <dbl>   <int>     <dbl>     <dbl>
1      2    24   32.0 1.27e- 1      26      498.  1.13e-88
2      1    50  530.  2.60e-81      28     3183.  0       
3      0    78 3713.  0             NA       NA  NA     

While Yukaghirs work with a CentralYakutia_LN base.

Code:

# A tibble: 3 x 5
  target            left                        weight     se     z
  <chr>             <chr>                        <dbl>  <dbl> <dbl>
1 Yukagir_Tundra.HO Russia_CentralYakutia_LN.SG  0.636 0.0571 11.1
2 Yukagir_Tundra.HO Tuvinian.HO                  0.147 0.0406  3.62
3 Yukagir_Tundra.HO Itelmen.HO                   0.217 0.0458  4.73

$rankdrop
# A tibble: 3 x 7
  f4rank   dof  chisq         p dofdiff chisqdiff   p_nested
   <int> <int>  <dbl>     <dbl>   <int>     <dbl>      <dbl>
1      2    23   27.0 2.54e-  1      25      378.  1.01e- 64
2      1    48  405.  5.60e- 58      27     1297.  2.44e-256
3      0    75 1701.  1.35e-305      NA       NA  NA       

Code:

$weights
# A tibble: 2 x 5
  target            left                        weight      se     z
  <chr>             <chr>                        <dbl>   <dbl> <dbl>
1 Yukagir_Forest.HO Russia_CentralYakutia_LN.SG  0.455 0.00658  69.1
2 Yukagir_Forest.HO Russian.HO                   0.545 0.00658  82.7

$rankdrop
# A tibble: 2 x 7
  f4rank   dof  chisq     p dofdiff chisqdiff p_nested
   <int> <int>  <dbl> <dbl>   <int>     <dbl>    <dbl>
1      1    24   30.6 0.167      26     6538.        0
2      0    50 6569.  0          NA       NA        NA

[Jaska]

Actually makes the most sense that they would have Turkic ancestry, the half million Yakuts nearby didn't come from nothing. Todzin are from Tuva, should be a good proxy for the original proto/Siberian Turkic speakers. Maybe even N-P43 comes from there?

Proto-Samoyedic was spoken in the Sayan Region, and some Turkic populations there are former Samoyeds. N-P43 is the shared paternal lineage of all Samoyedic populations, so it probably was already present in the Proto-Samoyedic speaking population. But if there are similarities between the Nganasans and the Dolgans or the Yakuts alone, these are most likely the result of admixture in the Northernmost Siberia.

[Kale]

I tried Mansi next, even the best non feasible p was only 0.046, best feasible model is Russia_Karelia_HG + Russia_Mezhovskaya.SG + Russia_CentralYakutia_LN.SG p=0.000125.

Russia_Mansi.DG
Andronovo_Kytmanovo.SG 0.391159 0.0159156 24.5770
RUS_Tyumen_NHG 0.139238 0.0171709 8.10891
RUS_Krasnoyarsk_Nefteprovod_BA.SG 0.469603 0.0106728 44.0000
Tail: 0.32

I kind of 'doubled up' on right pops just to get the std. err. tight and to tell the somewhat similar Siberian sources apart, but in the end it was really straightforward, Kra001 and nothing else was necessary. Mezhovskaya also works, as with just about any Sintashta clone.
right = c('Congo_Mbuti.DG', 'Anatolia_Barcin_N.SG', 'Anatolia_Boncuklu_N.SG', 'Iran_TepeAbdulHosein_N.SG', 'Iran_Wezmeh_N.SG', 'CHG.SG', 'Morocco_OUB002_Epipaleolithic.SG', 'Morocco_SKH001_MN.SG', 'Italy_GrottaContinenza_HG.SG', 'Bichon.SG', 'Sweden_StoraForvar_HG.SG', 'RUS_Arkhangelsk_HG.SG', 'Botai.SG', 'Afanasievo_KarasukIII.SG', 'Yamnaya_Kazakhstan_Karagash.SG', 'Andaman_100BP.SG', 'Jomon.SG', 'RUS_Irkutsk_Glazkovo_EN.SG', 'RUS_Irkutsk_Shamanka_EN.SG', 'RUS_Irkutsk_AngaraRiver_LN.SG', 'RUS_Irkutsk_Sokhter_LN.SG', 'RUS_Yakutia_Kamenka2_LN.SG', 'RUS_Yakutia_Kyordyughen_LN.SG', 'RUS_Primorsky_DevilsCave_N.SG', 'Peru_RioUncallane_1800BP.SG')
allsnps=TRUE

[Ebizur]

Actually makes the most sense that they would have Turkic ancestry, the half million Yakuts nearby didn't come from nothing. Todzin are from Tuva, should be a good proxy for the original proto/Siberian Turkic speakers. Maybe even N-P43 comes from there?

Proto-Samoyedic was spoken in the Sayan Region, and some Turkic populations there are former Samoyeds. N-P43 is the shared paternal lineage of all Samoyedic populations, so it probably was already present in the Proto-Samoyedic speaking population. But if there are similarities between the Nganasans and the Dolgans or the Yakuts alone, these are most likely the result of admixture in the Northernmost Siberia.

Only the N-P43=N-B523 (TMRCA 4,542 [95% CI 5,333 - 3,866] ybp according to FTDNA) > N-VL67 (TMRCA 4,397 [95% CI 5,225 - 3,696] ybp) > N-VL63/N-B478 (TMRCA 2,822 [95% CI 3,810 - 2,057] ybp) subclade of N-P43 has a significant presence among extant speakers of Samoyedic languages. N-B478 also has been observed in fairly large percentages of samples of Siberian Turkic-speaking ethnic groups (Khakas, Tuvans, Shors, Altaians, Dolgans, Yakuts, etc.), and it also has been observed among Mongols, Evenks, and Evens.

N-P43=N-B523 > N-VL67 > N-B525 (TMRCA 3,024 [95% CI 3,637 - 2,511] ybp according to FTDNA) is rare everywhere, but it has been found mainly among Central Asian Muslims and among people who have historically interacted with Central Asian Muslims and Turks. The most basal known branches of N-B525 have been found in Iran, Turkey, and Greece. Overall, it seems very "Turkish" (as in Turkey Turks/Oghuz Turks rather than "Turkic peoples" in general).

N-P43 > N-Y3196 (TMRCA 4,511 [95% CI 5,370 - 3,786] ybp according to FTDNA) is almost equally ancient as N-P43 as a whole. Prior to recent Russian colonization of Siberia, N-Y3196 seems to have been found within an area roughly delimited by Finland in the west, Khantia-Mansia in the east, the Arctic Ocean in the north, and Ukraine and Southern Russia in the south, exhibiting a positive correlation with latitude within this area. I would say that N-P43 as a whole would be better considered as a member of a hypothetical Proto-Uralic-speaking population than as a member of a specifically Proto-Samoyedic-speaking population, although the MRCA of N-B478 plausibly may have been a speaker of the Proto-Samoyedic language.