G25 models on VURers and other Uralics

**Zanzibar** · 03-03-2021, 02:45 AM

What do you think? The distance runs are not that good but give good rough estimates into the autosomal composition of each Uralic and VURer population. Hungarians and Estonians seem to be the least genetically Uralic here and have the highest EEF meanwhile Khanty, Mansi, Selkup and Nenets seem to be the most Uralic and have the least EEF admixture.

What I find very interesting is how very low Neolithic Farmer that Saamis which is even lower than the amount that VURers possessed, Saamis seem to be an almost isolated population despite being located in Northern Scandinavia. On the other hand, Saami_Kola despite being in Russia, actually possess higher EEF than the Saami samples which probably indicate recent admixture with other European populations.

Comparison between Saami and Saami_Kola: as you can see, Saami_Kola are genetically more European and less Mongoloid than the Saami

Now I decided to model VURers and other Uralics with neighboring populations and the results are surprising:

P.S.- FIN_Levanluhta_o is an ancient Germanic individual from Finland, Czech_Early_Slav is an ancient Slav from Czechia, Baltic_EST_BA is a Bronze Age individual from Estonia, Uyelgi is an ancient Uralic, Ancient_Finno_Ugric is made by using an ancient Finnic individual from Viking Age, Norway (VK2020_NOR_North_VA_o1), KAZ_Botai is a West Siberian Hunter Gatherer population who are a mix of ANE+EHG+East Asian, Altaian and Uzbek are there to detect if there is any Turkic admixture, Tajik_Yagnobi and Iranian_Zoroastrian are utilized for proxies of Iranic affinity

I decided to do the same thing with the Saami and Saami_Kola individuals: Seems like most of the non-Uralic admixture in Saamis comes from Baltics followed by Nordics (Levanluhta_o) meanwhile the non-Uralic ancestry in Saami_Kola seems to come from Balts, Scandinavians and Slavs.

**Zanzibar** · 03-03-2021, 06:56 AM

Here are how individual Udmurts, Maris, Chuvashs, Besermyans, Komis, Bashkirs score:

Here is how they score in another model that I simulated:

**Zanzibar** · 03-03-2021, 07:15 AM

Here are how individual Mansi, Khanty, Nenets, Selkups score:

Compare to another model I did:

**Zanzibar** · 03-03-2021, 09:27 AM

Here are the results of individual Finns, Karelians, Vepsians, Ingrians, Mordovians, Estonians, Russian_Pinega and Hungarians:

Here is another model of them:

**~~Komintasavalta~~** · 03-03-2021, 12:19 PM

I tried using all populations from your first model as targets, except I didn't use Tundra_Nentsi or Forest_Nentsi, because they are not included in the official datasheet, and I added Nganasans who were missing. I used all rows from the ancient averages sheet as sources, except rows containing "_o", "_contam", or "_low_res".

These were initially the sources that got the highest average percentage:

15.0 RUS_Krasnoyarsk_BA
11.6 FIN_Levanluhta_IA
5.8 Baltic_EST_BA
5.4 UKR_MBA
5.4 KAZ_Golden_Horde_Euro
4.9 RUS_Chalmny-Varre
4.0 VK2020_UKR_Shestovitsa_VA
3.7 RUS_Ingria_IA
3.4 Baltic_LTU_BA
2.3 Baltic_EST_MA
2.2 RUS_Yakutia_Ymyiakhtakh_LN
1.9 VK2020_SWE_Uppsala_VA
1.9 RUS_Samara_HG
1.7 RUS_Maykop
1.7 KAZ_Botai
1.6 MNG_Chandman_IA
1.5 MNG_SHU002
1.5 KAZ_Zevakinskiy_BA
1.4 CZE_Early_Slav
1.2 VK2020_RUS_Kurevanikha_VA
1.2 USA_colonial_period
1.2 KAZ_Tasbas_IA
1.1 DEU_LBK_KD
1.0 VK2020_POL_Sandomierz_VA
0.8 KAZ_Kimak
0.7 VK2020_RUS_Pskov_VA
0.7 RUS_Tyumen_HG

Next I did a PCA of the sources above with clustering:

I picked one popultaion from each cluster as source in a new model. I didn't even try to choose the sources so that they minimized the average distance. For example I picked KAZ_Golden_Horde_Euro from the Northern European cluster because it sounded the coolest.

The average distance was .020 for the first model, .022 for the second model, and .029 for the third model.

What's surprising is that Chuvashes, Maris, and Udmurts got such a high percentage of Chalmny Varre. Chalmny Varre is a 18th-19th century Saami cemetery on the Kola Peninsula. Maybe Udmurts and Besermyans have a lower percentage of Chalmny Varre partially because they have over 10% RUS_Tyumen_HG (WSHG).

Nganasans are modeled as 98% RUS_Krasnoyarsk_BA (which only consists of the single individual kra001 that Davidski has recently speculated might be Proto-Uralic). When I removed Chalmny Varre, Finnish_East got 7% Krasnoyarsk_BA, Udmurts got 19%, Saami got 24%, and Maris got 29%.

MNG_SHU002 is the main Mongoloid component in Tatar_Siberian, Tatar_Crimean_steppe, and Tatar_Lipka. However the main Mongoloid component is the more Uralic-like RUS_Krasnoyarsk_BA in Zabolotniye Tatars (swamp Tatars), who are considered to be turkified Uralics.

The proportion of Chalmny Varre is 4% in Estonians, 22% in Finns, 33% in Karelians, 37% in Vepsians, and 43% in Pinega Russians. They all have 0% of Krasnoyarsk_BA.

DEU_LBK_KD and RUS_Maykop are the two woggiest component. LBK (Linearbandkeramik) has the highest percentage in Hungarians as expected. RUS_Maykop is more common in the Volga-Ural region but absent among Baltic Finnic peoples and Saami. Out of modern population averages, it is the closest to Tajiks and churkas:

Distance to: RUS_Maykop
.066 Tajik_Rushan
.070 Tajik_Shugnan
.071 Darginian
.072 Tajik_Yagnobi
.077 Avar
.077 Lak
.078 Kubachinian
.079 Kaitag
.081 Tabasaran
.082 Tajik_Ishkashim
.098 Tajik
.099 Chechen
.105 Balkar
.106 Kumyk
.107 Ingushian
.109 Cherkes

You can generate a clustered heatmap of CSV data from Vahaduo's MULTI tab by running this in R:

Code:

library(pheatmap)

download.file("https://drive.google.com/uc?export=download&id=1wZr-UOve0KUKo_Qbgeo27m-CQncZWb8y","modernave")

t=read.csv("input-from-multi-tab-in-vahaduo.csv",header=T,row.names=1,check.names=F)
avedist=t[nrow(t),1]
t=t[-nrow(t),]
t=t[order(row.names(t)),]

t2=read.csv("modernave",header=T,row.names=1,check.names=F)
t3=t2[row.names(t2)%in%row.names(t),]
k=hclust(dist(t3))

row.names(t)=paste0(row.names(t)," (",sub("^0","",sprintf("%.3f",t[,1])),")")
t=t[-c(1)]

pheatmap(
  t,
  filename="/tmp/a.png",
  clustering_callback=function(...){c(k)},
  cluster_cols=F,
  legend=F,
  main=paste("Average distance:",sub("^0","",sprintf("%.3f",avedist))),
  cellwidth=16,
  treeheight_row=100,
  cellheight=16,
  fontsize=10,
  border_color=NA,
  display_numbers=T,
  number_format="%.0f",
  fontsize_number=8,
  number_color="black",
  breaks=seq(0,100,100/256),
  colorRampPalette(hex(HSV(c(210,210,90,60,40,20,0,0),c(0,.4,.6,.6,.6,.6,.6,.8),c(1,1,1,1,1,1,1,.6))))(256)
)

Here's another PCA with both my source and target populations. Now Tyumen_HG (WSHG) actually ended up clustering with Samara_HG (EHG). I chose a too small number of clusters, so Hungarians clustered together with Finns. PC3 differentiates WSHG from Nganasan, and PC4 differentiates MNG_SHU002 (a Mongoloid source of Turkic peoples) from RUS_Maykop (Tajik/Churka).

**Zanzibar** · 03-03-2021, 12:35 PM

Originally Posted by Komintasavalta

I tried using all populations from your first model as targets, except I didn't use Tundra_Nentsi or Forest_Nentsi, because they are not included in the official datasheet, and I added Nganasans who were missing. I used all rows from the ancient averages sheet as sources, except rows containing "_o", "_contam", or "_low_res".

These were initially the sources that got the highest average percentage:

15.0 RUS_Krasnoyarsk_BA
11.6 FIN_Levanluhta_IA
5.8 Baltic_EST_BA
5.4 UKR_MBA
5.4 KAZ_Golden_Horde_Euro
4.9 RUS_Chalmny-Varre
4.0 VK2020_UKR_Shestovitsa_VA
3.7 RUS_Ingria_IA
3.4 Baltic_LTU_BA
2.3 Baltic_EST_MA
2.2 RUS_Yakutia_Ymyiakhtakh_LN
1.9 VK2020_SWE_Uppsala_VA
1.9 RUS_Samara_HG
1.7 RUS_Maykop
1.7 KAZ_Botai
1.6 MNG_Chandman_IA
1.5 MNG_SHU002
1.5 KAZ_Zevakinskiy_BA
1.4 CZE_Early_Slav
1.2 VK2020_RUS_Kurevanikha_VA
1.2 USA_colonial_period
1.2 KAZ_Tasbas_IA
1.1 DEU_LBK_KD
1.0 VK2020_POL_Sandomierz_VA
0.8 KAZ_Kimak
0.7 VK2020_RUS_Pskov_VA
0.7 RUS_Tyumen_HG

Next I did a PCA of the sources above with clustering:

I picked one popultaion from each cluster as source in a new model. I didn't even try to choose the sources so that they minimized the average distance. For example I picked KAZ_Golden_Horde_Euro from the Northern European cluster because it sounded the coolest.

The average distance was .020 for the first model, .022 for the second model, and .029 for the third model.

What's surprising is that Chuvashes, Maris, and Udmurts got such a high percentage of Chalmny Varre. Chalmny Varre is a 18th-19th century Saami cemetery on the Kola Peninsula. Maybe Udmurts and Besermyans have a lower percentage of Chalmny Varre partially because they have over 10% RUS_Tyumen_HG (WSHG).

Nganasans are modeled as 98% RUS_Krasnoyarsk_BA (which only consists of the single individual kra001 that Davidski has recently speculated might be Proto-Uralic). When I removed Chalmny Varre, Finnish_East got 7% Krasnoyarsk_BA, Udmurts got 19%, Saami got 24%, and Maris got 29%.

MNG_SHU002 is the main Mongoloid component in Tatar_Siberian, Tatar_Crimean_steppe, and Tatar_Lipka. However the main Mongoloid component is the more Uralic-like RUS_Krasnoyarsk_BA in Zabolotniye Tatars (swamp Tatars), who are considered to be turkified Uralics.

The proportion of Chalmny Varre is 4% in Estonians, 22% in Finns, 33% in Karelians, 37% in Vepsians, and 43% in Pinega Russians. They all have 0% of Krasnoyarsk_BA.

DEU_LBK_KD and RUS_Maykop are the two woggiest component. LBK (Linearbandkeramik) has the highest percentage in Hungarians as expected. RUS_Maykop is more common in the Volga-Ural region but absent among Baltic Finnic peoples and Saami. Out of modern population averages, it is the closest to Tajiks and churkas:

Distance to: RUS_Maykop
.066 Tajik_Rushan
.070 Tajik_Shugnan
.071 Darginian
.072 Tajik_Yagnobi
.077 Avar
.077 Lak
.078 Kubachinian
.079 Kaitag
.081 Tabasaran
.082 Tajik_Ishkashim
.098 Tajik
.099 Chechen
.105 Balkar
.106 Kumyk
.107 Ingushian
.109 Cherkes

You can generate a clustered heatmap of CSV data from Vahaduo's MULTI tab by running this in R:

Code:

library(pheatmap)

download.file("https://drive.google.com/uc?export=download&id=1wZr-UOve0KUKo_Qbgeo27m-CQncZWb8y","modernave")

t=read.csv("input-from-multi-tab-in-vahaduo.csv",header=T,row.names=1,check.names=F)
avedist=t[nrow(t),1]
t=t[-nrow(t),]
t=t[order(row.names(t)),]

t2=read.csv("modernave",header=T,row.names=1,check.names=F)
t3=t2[row.names(t2)%in%row.names(t),]
k=hclust(dist(t3))

row.names(t)=paste0(row.names(t)," (",sub("^0","",sprintf("%.3f",t[,1])),")")
t=t[-c(1)]

pheatmap(
  t,
  filename="/tmp/a.png",
  clustering_callback=function(...){c(k)},
  cluster_cols=F,
  legend=F,
  main=paste("Average distance:",sub("^0","",sprintf("%.3f",avedist))),
  cellwidth=16,
  treeheight_row=100,
  cellheight=16,
  fontsize=10,
  border_color=NA,
  display_numbers=T,
  number_format="%.0f",
  fontsize_number=8,
  number_color="black",
  breaks=seq(0,100,100/256),
  colorRampPalette(hex(HSV(c(210,210,90,60,40,20,0,0),c(0,.4,.6,.6,.6,.6,.6,.8),c(1,1,1,1,1,1,1,.6))))(256)
)

Here's another PCA with both my source and target populations. Now Tyumen_HG (WSHG) actually ended up clustering with Samara_HG (EHG). I chose a too small number of clusters, so Hungarians clustered together with Finns. PC3 differentiates WSHG from Nganasan, and PC4 differentiates MNG_SHU002 (a Mongoloid source of Turkic peoples) from RUS_Maykop (Tajik/Churka).

Well, there are individual Forest_Nentsi and Tundra_Nentsi individual samples in G25 but they haven't been averaged yet, so I decided to do it myself. I will have to ask Lucas to include them as well into the modern averages spreadsheet. Here are the averages. You can include them into your runs:

Code:

Forest_Nentsi,0.067358893,-0.288337893,0.131682429,0.03657975,-0.1223635,-0.056027214,0.017541964,0.025770893,0.01321375,-0.014442179,0.078851179,0.001760929,0.011961857,-0.066845357,-0.022844643,-0.012894357,-0.004316571,0.006406786,0.010437357,-0.0099065,0.007170393,0.02140075,0.036366893,-0.007160857,-0.004896821

Tundra_Nentsi,0.069782231,-0.273255462,0.127408769,0.034064077,-0.116045308,-0.050779615,0.017155769,0.025703077,0.012586154,-0.013401462,0.069527231,0.001902077,0.010554923,-0.054773769,-0.018228231,-0.009842231,1E-05,0.005145308,0.006072154,-0.007859538,0.008648231,0.016293692,0.027787692,-0.002224615,-0.001381692

Very interesting. Thanks for the feedback! Well you can also use RUS_Shamanka_N or RUS_Devil's Gate for the Mongoloid ancestry of most Tatars.

Btw can you tried to find out how much EEF and CHG each of these Uralics possessed by using TUR_Barcin_N and GEO_CHG? You will have to remove the Chamny_Varre and Kaz_Golden_Horde_Euro components though as they also contain Neolithic and CHG (from Steppe admixture) affinity. Also add Baltic_LVA_HG to get a better distance run it seems.

Is DEU_LBK_KD an EEF component? I checked and they seem closest to Sardinians followed by other South Euros. While RUS_Maykop is closest to MENA.

**~~Komintasavalta~~** · 03-03-2021, 01:40 PM

Originally Posted by Joqool

Well, there are individual Forest_Nentsi and Tundra_Nentsi individual samples in G25 but they haven't been averaged yet, so I decided to do it myself. I will have to ask Lucas to include them as well into the modern averages spreadsheet.

Ok, they were only included on Lucas's website but not here: https://eurogenes.blogspot.com/2019/...obal25_12.html.

I used this to calculate population averages for Forest_Nentsi and Tundra_Nentsi, and I got the same result as you:

sed 's/:[^,]*//'|awk -F, '{n[$1]++;for(i=2;i<=NF;i++){a[$1][i]+=$i}}END{for(i in a){o=i;for(j=2;j<=NF;j++)o=o","a[i][j]/n[i];print o}}'

Originally Posted by Joqool

Well you can also use RUS_Shamanka_N or RUS_Devil's Gate for the Mongoloid ancestry of most Tatars.

Yeah or RUS_Lokomotiv_N. It's extremely close to Shamanka:

Distance to: RUS_Lokomotiv_N
.014 RUS_Shamanka_N
.020 RUS_Fofonovo_En
.027 RUS_Baikal_BA_o
.031 MNG_North_N
.033 MNG_Slab_Grave_EIA_1
.034 RUS_Yakutia_Meso
.036 MNG_EIA_3
.046 RUS_Yakutia_N

Originally Posted by Joqool

Btw can you tried to find out how much EEF and CHG each of these Uralics possessed by using TUR_Barcin_N and GEO_CHG? You will have to remove the Chamny_Varre and Kaz_Golden_Horde_Euro components though as they also contain Neolithic and CHG (from Steppe admixture) affinity. Also add Baltic_LVA_HG to get a better distance run it seems.

I tried using TUR_Pinarbasi_HG instead of TUR_Barcin_N. It increased the average distance by about .002 compared to TUR_Barcin_N, but I think it's cooler to use HGs.

I got about .001 lower average distance with NOR_N_HG than with Baltic_LVA_HG. If I included both in sources, Vahaduo only used NOR_N_HG in the MULTI tab.

MNG_SHU002:SHU002 is dated to about about 1212 CE. It gave me a lower average distance than older ancient sources like RUS_Shamanka_N or RUS_Lokomotiv_N. (Which makes sense because the Turkic expansion did not take place in the Neolitihic.)

Finnics had almost the same amount of GEO_CHG as VURers. When I tried replacing GEO_CHG with RUS_Maykop, the percentage in Finnics became clearly lower than in VURers.

The proportion of MNG_SHU002 (Mongoloid source of Turkic peoples) was 14% in Turkish_Southwest and 29% in Turkmen. Both of them also had fairly high WSHG.

Swamp Tatars still have about 3 times higher RUS_Krasnoyarsk_BA (kra001) than MNG_SHU002. Also Swamp Tatars cluster together with Uralics but other Siberian Tatars cluster with Bashkirs and Turkmen.

**Zanzibar** · 03-03-2021, 02:00 PM

Originally Posted by Komintasavalta

Ok, they were only included on Lucas's website but not here: https://eurogenes.blogspot.com/2019/...obal25_12.html.

I used this to calculate population averages for Forest_Nentsi and Tundra_Nentsi, and I got the same result as you:

sed 's/:[^,]*//'|awk -F, '{n[$1]++;for(i=2;i<=NF;i++){a[$1][i]+=$i}}END{for(i in a){o=i;for(j=2;j<=NF;j++)o=o","a[i][j]/n[i];print o}}'

Yeah or RUS_Lokomotiv_N. It's extremely close to Shamanka:

Distance to: RUS_Lokomotiv_N
.014 RUS_Shamanka_N
.020 RUS_Fofonovo_En
.027 RUS_Baikal_BA_o
.031 MNG_North_N
.033 MNG_Slab_Grave_EIA_1
.034 RUS_Yakutia_Meso
.036 MNG_EIA_3
.046 RUS_Yakutia_N

I tried using TUR_Pinarbasi_HG instead of TUR_Barcin_N. It increased the average distance by about .002 compared to TUR_Barcin_N, but I think it's cooler to use HGs.

I got about .001 lower average distance with NOR_N_HG than with Baltic_LVA_HG. If I included both in sources, Vahaduo only used NOR_N_HG in the MULTI tab.

MNG_SHU002:SHU002 is dated to about about 1212 CE. It gave me a lower average distance than older ancient sources like RUS_Shamanka_N or RUS_Lokomotiv_N. (Which makes sense because the Turkic expansion did not take place in the Neolitihic.)

Finnics had almost the same amount of GEO_CHG as VURers. When I tried replacing GEO_CHG with RUS_Maykop, the percentage in Finnics became clearly lower than in VURers.

The proportion of MNG_SHU002 (Mongoloid source of Turkic peoples) was 14% in Turkish_Southwest and 29% in Turkmen. Both of them also had fairly high WSHG.

Swamp Tatars still have about 3 times higher RUS_Krasnoyarsk_BA (kra001) than MNG_SHU002. Also Swamp Tatars cluster together with Uralics but other Siberian Tatars cluster with Bashkirs and Turkmen.

Actually I have noticed that it seems that Pinarbasi_HG slightly inflated the amount of EEF among Uralics so its better to utilize Barcin_N as a proxy instead. Also try adding Baltic Hunter Gatherers like Baltic_LVA_HG and ancient Uralics like RUS_Bolshoy_Oleni_Ostrov_o, it might help decrease the EEF.

Btw are you shocked that even Mari, Saami, Bashkir, Khanty, Mansi have higher EEF and CHG than expected? I believe the RUS_Maykop or CHG in Uralics come from Steppe admixture. Steppe peoples like Yamnaya were genetically half EHG and half wog (CHG) with some minor EEF wog as well.

I also forget to mention that there seem to be new Khants samples as well which were recently added. They also haven't been average, so I decided to average them. Can you also included them also into your run?

Code:

Khants,0.088877,-0.171285917,0.11184775,0.06112775,-0.079860917,-0.02424025,0.009772333,0.01648,-0.000238667,-0.0302815,0.051355333,-0.00370925,0.02143175,-0.069935333,-0.025832167,-0.01510425,-0.001075583,0.001224667,-0.003006333,-0.010702917,-0.002682667,0.019660833,0.021322,-0.0015565,-0.006196917

Must mentioned though that the Mari in G25 seems to suffers from an extreme genetic drift that's why the distance fit is pretty terrible.

Can you try running these Saami and Mari individuals?

Code:

Saami:GS000035025,0.103579,-0.04773,0.121433,0.083657,-0.018773,0.007251,0.005405,0.011307,0.00225,-0.033896,0.028093,-0.008243,0.021853,-0.017065,-0.008143,-0.008353,-0.015646,0.00114,-0.002137,-5e-04,0.019341,-0.003462,-0.006655,-0.001325,0.001557
Saami:GS000035026,0.112685,-0.019295,0.112759,0.077197,0.000615,0.003068,0.0094,0.010615,0.007976,-0.032438,0.022572,-0.007194,0.016799,-0.018166,-0.007872,0.005569,0.011343,0.001267,-0.005154,0.006003,0.020713,0.001607,-0.002711,-0.000723,0.000239
Saami:Saami001,0.118376,-0.005078,0.10484,0.077197,-0.001846,0.011156,0.01034,0.013615,-0.002045,-0.032985,0.018837,-0.008692,0.021407,-0.007432,-0.003664,-0.003182,-0.005476,-0.002154,-0.005028,0.008754,0.019466,0.000989,-0.008011,0.003615,-0.003712
Saami:saami1,0.103579,-0.035544,0.111251,0.077197,-0.010463,0.00753,0.01034,0.017076,0.004295,-0.034625,0.02517,-0.01139,0.021556,-0.015414,-0.008686,-0.008221,-0.0103,-0.000507,-0.00729,-0.001626,0.019091,0.000989,-0.004314,-0.001687,-0.002395
Saami:saami11,0.110408,-0.052808,0.113136,0.078166,-0.015387,0.008088,0.008695,0.014076,0.005727,-0.030433,0.031991,-0.003147,0.020069,-0.026974,-0.003122,0.001989,0.004042,-0.00038,-0.008547,-0.002501,0.017968,0.001978,0,0.004458,0.00467
Saami:saami12,0.10927,-0.01828,0.111251,0.079135,-0.01231,0.009482,0.011516,0.016615,-0.000409,-0.032074,0.024033,-0.01079,0.016501,-0.015964,-0.009636,0.01074,0.00665,-0.000633,-0.010307,0.01063,0.016471,0.004822,-0.00419,-0.000361,0.001197
Saami:saami13,0.114961,-0.031481,0.110496,0.083334,-0.00954,0.009761,0.012456,0.014999,0.003272,-0.033167,0.019649,-0.01139,0.011596,-0.024772,-0.006786,0.009679,0.00665,-0.002407,-0.003142,0.004377,0.012603,-0.001484,0.003328,0.006868,-0.003353
Saami:saami14,0.113823,-0.037575,0.106725,0.084626,-0.006155,0.006414,0.00423,0.011076,0.001227,-0.035354,0.028418,-0.008842,0.021258,-0.020506,-0.004479,0.004243,0.004824,-0.004054,-0.006159,0.006753,0.010606,0.000371,0.000246,0.008796,-0.001796
Saami:saami2,0.108132,-0.052808,0.11653,0.078489,-0.022466,0.005578,0.005875,0.008307,0.006545,-0.03262,0.020948,-0.006294,0.017542,-0.020506,-0.001357,-0.008618,-0.014212,-0.00076,-0.003142,0.008879,0.015722,0.006554,-0.000246,-0.002048,0.00012
Saami:saami3,0.111547,-0.027419,0.108611,0.076874,-0.011079,0.004462,0.01081,0.018692,0.003886,-0.034625,0.026956,-0.006594,0.019029,-0.023946,-0.00475,0.01074,0.014733,-0.002787,-0.002765,0.004127,0.008111,0.001237,-0.000986,0.004097,0.000359
Saami:saami4,0.106994,-0.004062,0.115776,0.087533,-0.003077,0.01255,0.006345,0.020538,0.003068,-0.031527,0.018025,-0.005845,0.017988,-0.016377,-0.004479,0.009812,0.009648,-0.010642,-0.01345,0.006003,0.021712,-0.001113,-0.002835,0.001205,0.004311
Saami:saami5,0.105855,-0.035544,0.112382,0.077843,-0.006463,0.00753,0.00517,0.010846,0.005727,-0.026242,0.02176,-0.008093,0.01888,-0.013625,0.005157,-0.013392,-0.028815,0.003927,-0.002011,-0.002501,0.021462,0.000495,-0.001109,-0.001807,0.00479
Saami:saami6,0.117238,-0.01828,0.107479,0.07752,-0.002154,0.016455,0.00705,0.016845,0.003068,-0.026242,0.0177,-0.005245,0.017096,-0.02257,-0.000679,0.011138,0.009779,0.00114,-0.00088,0.005503,0.006114,0.000124,-0.000616,0.010001,0.001676
Saami:saami7,0.112685,-0.031481,0.115776,0.083334,-0.012925,0.005857,0.004935,0.007615,-0.0045,-0.035718,0.022572,-0.006744,0.012636,-0.028488,-0.002714,0.008353,0.011213,-0.002027,-0.010056,0,0.012977,0.00272,-0.006162,0.000361,-0.003832
Saami:saami8,0.113823,-0.037575,0.108611,0.078812,-0.008617,0.005299,0.005875,0.01523,0.0045,-0.033349,0.013153,-0.004946,0.017988,-0.021194,-0.001629,0.002254,0.006258,-0.003674,-0.007165,0.009505,0.009982,0.003586,-0.001109,0.00976,0.003353
Saami:saami9,0.117238,-0.028435,0.110496,0.075582,-0.008617,0.012271,0.011751,0.014307,0.002863,-0.028247,0.016726,-0.006145,0.016947,-0.024222,-0.001221,0.010342,0.007823,-0.001267,-0.008547,0.008504,0.020713,0.002102,-0.005669,0.001325,-0.001676

Code:

Mari:GRC11056593,0.10927,-0.034528,0.09164,0.063631,-0.024312,-0.004183,0.009165,0.01823,-0.00409,-0.036994,0.025982,-0.013488,0.025421,-0.037296,-0.038002,-0.029037,-0.003912,-0.006208,-0.037835,-0.037018,0.006489,0.008408,-0.038823,0.010122,0.004311
Mari:GRC11056594,0.106994,-0.046714,0.093149,0.072675,-0.02739,0.000279,0.010105,0.010846,-0.005522,-0.035718,0.020461,-0.014237,0.031665,-0.046379,-0.039223,-0.02254,0.013821,-0.017483,-0.046131,-0.038518,0.013351,0.002968,-0.05349,0.009399,0.000239
Mari:GRC11056598,0.101303,-0.041637,0.090132,0.057817,-0.02739,0.008367,0.013866,0.013153,-0.010022,-0.046106,0.023709,-0.009591,0.030921,-0.035644,-0.029044,-0.024662,-0.006389,-0.012669,-0.037458,-0.034767,0.009483,0.003091,-0.046588,-0.003615,-0.00455
Mari:GRC11056599,0.097888,-0.052808,0.101823,0.065892,-0.031083,0.002789,0.018331,0.01823,-0.003477,-0.040456,0.024521,-0.009142,0.031367,-0.032479,-0.037052,-0.013922,0.012256,-0.008361,-0.042486,-0.026888,0.017469,0.001237,-0.042274,0.018677,-0.002634
Mari:mari1,0.105855,-0.060932,0.094657,0.059755,-0.029236,0.000279,0.011986,0.016615,-0.007363,-0.039545,0.031179,-0.013638,0.033003,-0.040048,-0.030537,-0.028772,-0.012647,-0.009755,-0.0406,-0.033016,0.014974,0.00507,-0.043876,0.008676,-0.005149
Mari:mari2,0.097888,-0.043668,0.083721,0.061047,-0.032929,-0.002789,0.011751,0.017076,0.001636,-0.040456,0.029392,-0.014687,0.034192,-0.035231,-0.035423,-0.027844,-0.009257,-0.010642,-0.038966,-0.030765,0.017594,0.003957,-0.054968,0.015544,0.000838
Mari:mari3,0.105855,-0.053823,0.096166,0.059109,-0.026159,0.005299,0.009635,0.018692,-0.007363,-0.038634,0.025982,-0.013638,0.0333,-0.038121,-0.032166,-0.02254,-0.013821,-0.007221,-0.037709,-0.039519,0.01123,0.009274,-0.047327,0.010845,0.002036
Mari:mari4,0.099026,-0.04773,0.097297,0.067507,-0.032006,-0.000837,0.011986,0.020538,0.000614,-0.036812,0.02858,-0.014687,0.030327,-0.039773,-0.038409,-0.022673,0.009909,-0.004181,-0.039469,-0.025512,0.018218,0.004204,-0.051394,0.008917,-0.00491
Mari:mari5,0.097888,-0.04773,0.092395,0.060078,-0.028005,0.001394,0.012456,0.019153,-0.006136,-0.041185,0.026307,-0.019483,0.041476,-0.044177,-0.037866,-0.016971,0.000913,-0.008361,-0.037081,-0.029889,0.01959,0.003957,-0.055831,0.010363,-0.003712

**Lemminkäinen** · 03-03-2021, 02:50 PM

Joqool, your results look interesting and decent. There are some results which will receive undeserved criticism from the Finnish side though.

**~~Komintasavalta~~** · 03-03-2021, 02:51 PM

I got bad fits for the new Mari samples as well. (The new samples have GRC in the name. The five samples with names like Mari:mari1 are part of the regular G25 datasheet.)

I still got a .002 lower average distance with NOR_N_HG than with Baltic_LVA_HG, but it's probably because there's so many Saami. Maris still get a lot of Samara_HG and very little NOR_N_HG.

The distance between new and old Khanty was .014. The distance between the new and old Maris was also only .018.

I also tried creating a model optimized just for Maris. I first added all lines from the ancient averages datasheet to the source tab. Then I made a new model using just the 6 sources with the highest average percentage. The average distance was still .082, but it's insane how much Levänluhta Maris get.

When I removed Levänluhta from the sources, its place was taken by Chalmny Varre. When I also removed Chalmny Varre, the main components in Maris turned into RUS_Krasnoyarsk_BA and Yamnaya_UKR.