PDA

View Full Version : Turkish Eurogenes Basal-Rich K7 & Globe10 Data Request



DMXX
08-22-2017, 01:37 PM
I've been dabbling with some nMonte ancients modelling for Iranians over the past few days and wanted to produce some decent results for modern Turks as well. However, the Basal-Rich K7 public spreadsheet only features "Turkish" and "Turkish Trabzon" populations. If there's enough traffic here, I'm intending to share the results on my research blog (http://vaedhya.blogspot.co.uk/), including maps and some basic statistics.

I'd appreciate receiving results for any individuals (including related) who are entirely of ancestry from the modern Turkish state. If anyone shares results for Turks with ancestry from outside the modern Turkish state, including Turkish-related groups (i.e. Pomak, Meshkhetian Turks), I'll happily run their data through my setup, but I won't include their data in the aforementioned blog entry.



[1] "distance%=1.0938 / distance=0.010938"

Turkish:average

Levant_N:average 48.8
Iran_N:average 25.0
Yamnaya:average 15.1
Altai_IA:average 11.1
Paniya:PNYD3 0.0
Sintashta:average 0.0
SSA_Excess 0.0
Loschbour:Loschbour 0.0


[1] "distance%=1.3495 / distance=0.013495"

Turkish_Trabzon:average

Levant_N:average 61.9
Iran_N:average 20.9
Yamnaya:average 17.3
Paniya:PNYD3 0.0
Sintashta:average 0.0
SSA_Excess 0.0
Loschbour:Loschbour 0.0
Altai_IA:average 0.0


A few comments:
- Iran_N + Levant_N + WHG combo to mirror Anatolian_BA (per Lazaridis et al.). Also accounts for any surplus of any component. I tried a synthetic, but it didn't produce fits anywhere near as good.
- Altai_IA is there for obvious reasons.
- EMBA and LNBA steppe samples (Yamnaya and Sintashta respectively) consolidated into averages.
- Paniya and SSA present for additional atypical admixture.

I tried this generic setup with Armenians and Iranians. The results were as anticipated. Armenians lacked East Eurasian admixture and possessed roughly equivalent Yamnaya to Turks (see below). This setup isn't ideal for Iranians (ideal looks like Iran_ChL + Iran_N + LNBA + East Eurasian, both for nMonte and qpAdm).



[1] "distance%=1.4388 / distance=0.014388"

Armenian:average

Levant_N:average 55.8
Iran_N:average 29.1
Yamnaya:average 15.2
Paniya:PNYD3 0.0
Sintashta:average 0.0
SSA_Excess 0.0
Loschbour:Loschbour 0.0
Altai_IA:average 0.0


Thank you!

alhan
08-22-2017, 03:15 PM
Dear DMXX,
I'd like to volunteer for your research,having all known recent ancestry from Turkey. However, I might need instructions "For dummies".

Do you just need results of Eurogenes_ANE K7?

DMXX
08-22-2017, 03:18 PM
All I need is the K7 and Globe10 personal results from one of the Eurogenes runs from last year.

This information can't be retrieved from an online calculator. Davidski directly sends the data to participants after receiving an appropriate fee.

I'd prefer Globe10 over K7, in case volunteers were wondering!

eolien
08-22-2017, 03:33 PM
Does it matter if such a sample has origins in Balkans and had turkish speaking ancestors (obviously not bosnians,or albanians etc)?

DMXX
08-22-2017, 03:39 PM
^ If I'm going to be creating plots showing the variation of various components across Anatolia, most certainly, since they're not originally from Anatolia.

I've offered to run our Turkish members irrespective of background through this nMonte setup irrespective, however.

alhan
08-22-2017, 03:51 PM
All I need is the K7 and Globe10 personal results from one of the Eurogenes runs from last year.

This information can't be retrieved from an online calculator. Davidski directly sends the data to participants after receiving an appropriate fee

I've had a look at the blog for K7 and Globe10. It seems complicated to me. Even if I was able to follow all the steps, I doubt that the results are going to be beyond my comprehension.

Need more reading first..

DMXX
08-22-2017, 04:57 PM
^ Have you participated in K7/G10? If you have, all I'd require are your results.

I've located the Turkish sub-groups from Davidski's older spreadsheet. Will share the results here shortly for everyone's viewing pleasure. Performing these with globe10.

[Edit]: Made some changes. The problems I had with the K7 Anatolian aDNA samples doesn't seem to occur with G10. Introducing several Anatolian aDNA averages and adding Nganasan+Han for East Eurasian (some of the Altai_IA scores were beginning to exceed 30%; Turkish LNBA heritage is getting absorbed by that).

DMXX
08-22-2017, 07:32 PM
After plenty of testing...



[1] "distance%=0.2182 / distance=0.002182"

Turkish_Balikesir

Tepecik_Ciftlik_N:avg 50.05
Yamnaya_Samara:avg 19.85
Kotias 10.05
Nganasan 7.30
Iran_N:avg 7.05
Han 5.70

[1] "distance%=0.2177 / distance=0.002177"

Turkish_Istanbul

Barcin_N:avg 28.20
Tepecik_Ciftlik_N:avg 18.50
Yamnaya_Samara:avg 15.95
Kotias 13.95
Iran_N:avg 13.40
Nganasan 5.55
Han 4.45

[1] "distance%=0.2467 / distance=0.002467"

Turkish_Kayseri

Tepecik_Ciftlik_N:avg 49.85
Sintashta:avg 22.75
Barcin_N:avg 14.25
Iran_N:avg 11.15
Nganasan 0.80
Han 0.75
Yamnaya_Kalmykia:avg 0.30
Yoruba 0.15

[1] "distance%=0.3439 / distance=0.003439"

Turkish_Trabzon

Tepecik_Ciftlik_N:avg 35.6
Kotias 28.2
Barcin_N:avg 18.6
Iran_N:avg 11.8
Yamnaya_Samara:avg 2.6
Nganasan 2.3
Han 0.8


Just waiting for Aydin and Adana samples to come out.

[Edit]: All done:



[1] "distance%=0.2905 / distance=0.002905"

Turkish_Aydin

Tepecik_Ciftlik_N:avg 51.55
Iran_N:avg 14.70
Yamnaya_Kalmykia:avg 14.10
Kotias 6.50
Nganasan 5.55
Barcin_N:avg 4.15
Han 3.45

[1] "distance%=0.3609 / distance=0.003609"

Turkish_Adana

Barcin_N:avg 26.75
Tepecik_Ciftlik_N:avg 25.85
Iran_N:avg 22.40
Yamnaya_Kalmykia:avg 12.15
Kotias 5.60
Nganasan 5.05
Han 2.20
Yamnaya_Samara:avg 0.00
Sintashta:avg 0.00
Paniya 0.00
Yoruba 0.00
Boncuklu_N:avg 0.00


My suspicion?

I don't think the "Iran_N" is truly from Iran in most of the models (at least in an informative sense). Time-predicated stratifying of Anatolian samples across the region hasn't happened just yet. I suspect much of the "Iran_N", particularly in the eastern and southern half of Turkey, is a local variant of Anatolian_ChL/BA that happens to have an excess of this component relative to what has been defined by Lazaridis et al.

I do, however, think some of it is real and was introduced by the Oghuz Turks (and those that followed them) via Khorasan. Combine that with Nganasan, Han and some portion of the West Eurasian steppe (a proportion of the excess relative to Armenians?), it looks reasonable to assume the waves contributed at least 20%.

eolien
08-23-2017, 08:51 AM
So, none of the turkish regional samples pick up ALTAI_IA but the so called 'turkish' does?

DMXX
08-23-2017, 09:15 AM
As stated in post #8, I removed Altai_IA and replaced it with Nganasan and Han. I made the switch specifically to clear up the degree of East Eurasian admixture. Altai_IA and Karasuk are both predominantly LNBA steppe with significant East Eurasian admixture. Ergo, an Altai_IA component will be particularly high among the Balikesir and Aydin samples.

I don't anticipate there was any significant Sintashta-like/"pure" LNBA steppe ancestry in Anatolia (odds are reasonable that any Cimmerian and Scythian input would have been EMBA, going by the Sarmatian samples). I included that in the first run to see whether any patterns emerged. It only popped up among the Kayseri sample.

This run was useful as it gives us an approximate range for the EMBA steppe (~2-20%) and East Eurasian (~1.5-13%) admixture among modern Turks with Anatolian roots. It served its' purpose.

In the next run, I'll be dropping Nganasan, Han, Sintashta and Boncoklu and adding Altai_IA back in (or Karasuk_outlier, depending on which is more East Eurasian). That should also address the slight overfitting seen above.

It's a work-in-progress, but I figured everyone would appreciate seeing the process from start to finish.

[Edit]: I've configured a pretty neat pan-West Asian setup that would theoretically work for South and North Caucasus pops, Armenians, Kurds, Iranians and Assyrians. Depending on the results, I might share the results. If nothing more comes of it, you know I'm not satisfied with the statistical fits.

DMXX
08-23-2017, 05:16 PM
Used Karasuk_outlier instead of Altai_IA. The virtue of using Yamnaya or Sintashta as the only steppe samples looks clear, here. The majority of the steppe is swallowed up by K_o.



[1] "distance%=0.3727 / distance=0.003727"

Turkish_Balikesir

Tepecik_Ciftlik_N:avg 49.95
Karasuk_outlier 23.50
Kotias 20.55
Yamnaya_Samara:avg 4.30
Paniya 1.55
Yoruba 0.15

[1] "distance%=0.4172 / distance=0.004172"

Turkish_Adana

Tepecik_Ciftlik_N:avg 54.05
Kotias 26.40
Karasuk_outlier 14.50
Paniya 5.05
Yamnaya_Samara:avg 0.00
Yoruba 0.00

[1] "distance%=0.3212 / distance=0.003212"

Turkish_Istanbul

Tepecik_Ciftlik_N:avg 49.4
Kotias 24.8
Karasuk_outlier 17.1
Yamnaya_Samara:avg 5.6
Paniya 3.1
Yoruba 0.0

[1] "distance%=0.3266 / distance=0.003266"

Turkish_Aydin

Tepecik_Ciftlik_N:avg 55.5
Kotias 23.8
Karasuk_outlier 17.4
Paniya 3.3
Yoruba 0.1
Yamnaya_Samara:avg 0.0

[1] "distance%=0.3873 / distance=0.003873"

Turkish_Trabzon

Tepecik_Ciftlik_N:avg 55.45
Kotias 37.10
Karasuk_outlier 4.95
Paniya 2.50
Yamnaya_Samara:avg 0.00
Yoruba 0.00

[1] "distance%=0.2833 / distance=0.002833"

Turkish_Kayseri

Tepecik_Ciftlik_N:avg 73.15
Yamnaya_Samara:avg 19.15
Paniya 2.95
Kotias 2.75
Karasuk_outlier 1.40
Yoruba 0.60

Eetion
08-24-2017, 07:40 AM
I'd appreciate receiving results for any individuals (including related) who are entirely of ancestry from the modern Turkish state. If anyone shares results for Turks with ancestry from outside the modern Turkish state, including Turkish-related groups (i.e. Pomak, Meshkhetian Turks), I'll happily run their data through my setup, but I won't include their data in the aforementioned blog entry.


Turkey Population: 80million (2017)
People who has Balkan ancestry 15-20 million
Kurmancı, Zaza, Muslim Georgion, Arab population and other 15-20 million

I give you respect. Selecting experimental DNA group is main thing in these kind of researchs.

So please keep going to me selective like this. ;)

DMXX
08-24-2017, 10:22 AM
^ Just expanding on the point below.

If the objective is to properly flesh out the ancestral profile of various groups within a nation as diverse as Turkey, specificity is an unavoidable principle (as is the case anyway in pop. genetics, epidemiology and a myriad of other related disciplines). That shouldn't be mistaken for discrimination (though it often is by laypeople). The numbers posted are free of value judgement.

There's a lot of controversy and heightened tensions surrounding the topic of ancestry in places like Turkey. Said tensions are exacerbated in other discussion forums online (if some viewers are also active in those places, any reservations felt upon viewing this thread are partly a consequence of their voluntary experience). This forum was founded as a rational break from such venues and a place where reason and empiricism prevail. That is precisely why I'm sharing my outputs as I went along. I certainly would benefit from member feedback and discussion.

Adopting an unspecific approach results in outputs which obscure internal diversity. I remember many of us were unhappy with the "Turkish_D" sample from the Dodecad project. We are thankfully beyond that point now.

Just to reiterate the point - I'm happy to share results for Turks with non-Anatolian ancestry. Just wanted to try and further the sample-set for Turkey beyond the six we currently have (Istanbul, Trabzon, Aydin, Adana, Balikesir, Kayser).

Alkaevli
08-26-2017, 06:12 PM
I'd also like to volunteer, unfortunately my father has not participated in K7/G10.

By the way, there is something unusual with Kayseri sample, Turk_Kayseri scores 6-8% East Eurasian admixture on gedmatch's calculators, but in your nMonte modellings even Turk_Trabzon's East Eurasian admixture (Nganasan + Han) seems to surpass that of Kayseri. Are your samples consisted of averages or random individuals from Turk_Aydin, Turk_Kayseri, Turk_Balikesir etc.? Perhaps that's an outlier/atypical individual from Turk_Kayseri.

Thanks for this thread.

DMXX
08-27-2017, 05:15 AM
By the way, there is something unusual with Kayseri sample, Turk_Kayseri scores 6-8% East Eurasian admixture on gedmatch's calculators, but in your nMonte modellings even Turk_Trabzon's East Eurasian admixture (Nganasan + Han) seems to surpass that of Kayseri. Are your samples consisted of averages or random individuals from Turk_Aydin, Turk_Kayseri, Turk_Balikesir etc.? Perhaps that's an outlier/atypical individual from Turk_Kayseri.

Yes, these are regional averages from Globe10 data.

Davidski produced a K7 run that includes individual Turkish data the other day:

https://docs.google.com/spreadsheets/d/1qYz-SsRykJEnlynp4MnQpjDETr4a0iZAZcAUJ8X2oT4/edit#gid=546161668

Eyeballing the total East Eurasian admixture among each Kayseri sample:



2%
7%
6.5%
3.5%
<1%
<1%
0%
<1%
0%
0%


Looks like the 1.5-2% average value was correct. Most of the individual Kayseri Turks are <1% or 0% East Eurasian.

Alkaevli
08-27-2017, 12:57 PM
I still think there is some mistake

Dienekes had posted individual data of Kayseri a few years ago. http://dienekes.blogspot.com.tr/2012/02/first-look-at-turkish-and-kyrgyz-data.html

The sample consists of 23 individuals and scores ~7% East Eurasian admixture on average.


Individiual data - Dodecad K12b values - http://dienekes.blogspot.com.tr/2012/02/first-look-at-turkish-and-kyrgyz-data.html?showComment=1329589827502#c2217395344610 213463

Kayseri 24032 17.9 4.1 1.1 0.0 12.9 7.6 0.9 0.0 12.9 3.0 39.6 0.0
Kayseri 23405 15.7 3.5 0.0 0.1 12.1 9.6 1.7 0.0 11.1 4.4 42.0 0.0
Kayseri 23469 8.0 1.9 0.0 0.5 22.1 24.5 0.0 0.0 10.9 4.0 28.2 0.0
Kayseri 23443 14.5 5.4 0.8 0.2 13.6 10.4 0.4 0.0 11.2 2.9 40.6 0.0
Kayseri 24266 14.9 4.8 1.1 0.6 13.0 9.6 0.6 0.0 13.0 1.2 41.0 0.0
Kayseri 24392 14.7 2.0 1.2 1.9 13.9 10.1 1.3 0.0 11.1 2.1 41.6 0.0
Kayseri 24276 15.6 2.9 0.6 1.6 14.8 5.5 0.3 0.0 12.4 1.8 44.5 0.0
Kayseri 24402 14.4 3.4 0.5 0.0 16.6 10.1 1.8 0.0 12.1 3.0 38.0 0.0
Kayseri 24166 15.0 3.4 0.0 0.0 11.5 9.6 0.2 0.0 12.8 4.5 43.1 0.0
Kayseri 24075 16.4 1.0 0.0 0.6 14.2 8.3 1.4 0.0 11.2 2.7 44.1 0.0
Kayseri 23967 15.6 4.9 1.0 0.0 12.4 9.8 0.0 0.1 10.8 4.2 41.2 0.0
Kayseri 23748 8.2 3.1 0.5 1.2 22.9 26.8 1.4 0.0 6.8 0.0 29.0 0.0
Kayseri 23827 13.2 3.9 1.0 1.4 14.8 9.3 1.7 0.0 11.0 2.3 41.4 0.0
Kayseri 24229 20.5 2.4 0.0 0.5 7.3 16.4 0.7 0.0 0.6 1.5 50.0 0.0
Kayseri 24424 15.3 2.8 0.7 0.8 14.2 8.2 2.6 0.0 11.2 2.1 42.1 0.0
Kayseri 23271 14.7 2.2 0.0 0.7 13.7 7.9 0.8 0.0 12.5 1.0 46.5 0.0
Kayseri 23988 22.4 2.7 1.0 0.4 8.6 8.5 2.2 0.0 11.6 0.7 41.8 0.0
Kayseri 23892 13.4 2.9 1.2 0.5 13.0 9.7 1.8 0.0 12.4 4.2 41.0 0.0
Kayseri 23242 14.0 3.4 2.2 0.0 15.0 9.9 0.9 0.0 9.7 4.2 40.6 0.0
Kayseri 23547 14.5 2.5 0.3 0.0 16.3 8.8 2.4 0.0 13.3 5.6 36.4 0.0
Kayseri 23315 14.9 4.3 0.4 2.2 13.0 8.4 0.0 0.0 12.4 4.7 39.6 0.0
Kayseri 23549 16.0 4.4 0.3 0.3 10.8 8.6 0.2 0.3 14.5 4.7 39.9 0.0
Kayseri 23989 16.2 3.6 2.7 1.2 9.0 11.3 2.4 0.0 13.5 1.9 38.1 0.0

Averages
https://i.hizliresim.com/6XAbOP.jpg (https://hizliresim.com/6XAbOP)



The results above are in line with Gedmatch's population averages, East Eurasian components are highlighted in red.
https://i.hizliresim.com/0BMAMW.png (https://hizliresim.com/0BMAMW)
https://i.hizliresim.com/kWNRNm.png (https://hizliresim.com/kWNRNm)

MDLP K23b
https://i.hizliresim.com/mkYZYV.png (https://hizliresim.com/mkYZYV)

Alkaevli
08-27-2017, 12:58 PM
The total East Eurasian admixture among each Kayseri individiual; Dodecad K12b / Steppe_EBA_K7 https://i.hizliresim.com/WQMy7q.png (https://hizliresim.com/WQMy7q) I still think that there might be some confusion with those labels in the spreadsheet, there is simply no correlation between Dodecad k12b and Steppe_EBA_K7 results, of course there is nothing you can do about it.

DMXX
08-27-2017, 01:54 PM
Probably best to ask Davidski about that, as he generated the data. Internally at least, the Eurogenes-derived data are consistent with another. Why there's a discrepancy with the other calculators is a good question.