Log in

View Full Version : Update on Balkan Turk Samples in G25



Kaspias
07-21-2020, 11:13 AM
I asked an update on Turkish_Balkans samples recently.

Changelog:


Correction of names of the current samples.
2 New samples from Komotini, 2 new samples from Kardzhali, 1 new sample from Thessaloniki, 1 new sample from Targovishte, 1 new sample from Razgrad.
Turkish_Balkans component distributed into Turkish_Deliorman and Turkish_Rumeli. Deliorman covers Northeastern/Northern Bulgaria while Rumeli covers Thrace and Macedonia.



Coordinates can be found both in official Eurogenes spreadsheet and here. Lukasz's Vahaduo needs an update.


SCALED
Turkish_Deliorman:Turkish_Silistre1,0.10927,0.0893 67,0.020365,0.002584,0.010156,0.000279,0.004935,0. 000231,-0.007567,0.001458,-0.009906,-0.005545,0.000297,0.015001,-0.009093,0,0.009518,0.002787,0.001508,-0.005628,-0.010981,0.002844,-0.000986,-0.001807,-0.001197
Turkish_Deliorman:Turkish_Silistre2,0.117238,0.101 553,0.021119,0.006137,0.015387,0.004183,0.008695,0 .005538,-0.006545,-0.007472,-0.002598,-0.000899,0.008176,0.007569,-0.005022,-0.000398,0.004694,0.001014,0.002263,0.006128,-0.00287,0.003339,-0.000246,0.003615,0.001078
Turkish_Deliorman:Turkish_Sumnu1,0.111547,0.105615 ,0.012068,-0.005491,0.005539,-0.001394,0.00846,0.002308,-0.007772,-0.000729,-0.00065,-0.004046,-0.003568,0.014588,-0.010043,-0.01432,-0.009648,0.001774,0.002765,-0.007504,-0.012852,-0.000618,-0.000739,-0.004458,-0.002036
Turkish_Deliorman:Turkish_Varna1,0.10927,0.101553, 0.012445,-0.002261,0.000615,0.006136,0.00564,0.008077,-0.005522,0.002916,-0.002111,-0.002398,0.001487,0.003716,-0.005157,0.010342,0.012778,0.000887,-0.001006,-0.002376,-0.007112,-0.000618,0.000246,0.007471,-0.007424
Turkish_Deliorman:Turkish_Eskicuma1,0.111547,0.092 413,0.020365,0.00323,0.013541,0.003626,0.00235,0.0 03231,-0.003068,0.002916,-0.001137,-0.007194,0.013974,-0.000826,-0.006107,0.003447,0.013821,0.005194,0.003645,0.000 125,0.007861,-0.005688,-0.001602,0.003133,0.000239
Turkish_Deliorman:Turkish_Razgrad1,0.113823,0.1167 86,0.023004,-0.004522,0.009848,0.001116,0.00282,0.011769,-0.002659,-0.005832,0.003735,-0.003597,0.001189,0.007432,-0.010315,-0.012331,-0.008345,0.000253,0.004651,-0.006378,-0.003993,-0.00371,-0.005793,-0.000482,0.006347
Turkish_Deliorman:Turkish_Razgrad2,0.10927,0.10967 7,0.020742,0,0.020004,0.005857,0.00423,0.00923,-0.007976,0.004556,0.002923,-0.002248,-0.000595,0.00867,-0.006243,-0.009281,-0.007171,0.001394,0.006788,-0.003627,-0.007237,-0.002844,0.000246,0.002771,0.000838
Turkish_Deliorman:Turkish_Razgrad3,0.10927,0.11069 3,0.02489,-0.000646,0.008001,-0.006693,0.00423,0.013153,-0.009204,-0.006196,-0.002111,-0.002398,0.003568,0.006881,-0.016829,-0.006099,0.002999,0.001647,-0.001885,-0.00025,-0.001872,-0.001978,-0.00037,-0.003856,0.000718
AVG
Turkish_Deliorman,0.111404375,0.103457125,0.019374 75,-0.000121125,0.010386375,0.00163875,0.00517,0.00669 2125,-0.006289125,-0.001047875,-0.001481875,-0.003540625,0.003066,0.007878875,-0.008601125,-0.00358,0.00233075,0.00186875,0.002341125,-0.00243875,-0.004882,-0.001159125,-0.0011555,0.000798375,-0.000179625

Turkish_Rumeli:Turkish_Selanik1,0.104717,0.12491,0 .006411,-0.00969,0.008001,-0.01004,0.00423,-0.003461,-0.002045,0.00492,0.009094,0.003447,-0.001041,0.005505,-0.007057,-0.000398,0.002217,0.008488,0.008296,-0.001751,-0.006613,0.002968,-0.001602,-0.004217,0.003233
Turkish_Rumeli:Turkish_Kilkis1,0.118376,0.110693,0 .008674,-0.010982,0.005539,0,0.00235,0.004384,-0.001432,-0.000182,0.002273,-0.003747,0.001784,0.006055,-0.018322,0.008751,0.020601,-0.009502,-0.000503,0.003001,-0.009608,0.004204,0.001479,0.001205,0.000239
Turkish_Rumeli:Turkish_Kircaali1,0.105855,0.087336 ,0.024513,-0.001938,0.00677,-0.011992,-0.001175,0.001846,-0.005318,0.003645,-0.004709,0.000599,0.003419,-0.004542,-0.011265,0.008751,0.014342,0.000887,0.002891,0.009 254,-0.004866,0.002102,0.003081,0.002169,0.005987
Turkish_Rumeli:Turkish_Kircaali2,0.113823,0.091398 ,0.010936,0.000323,0.011694,0,0.004935,-0.001385,-0.005318,0.001093,-0.003248,-3e-04,0.008771,0.001376,-0.007329,0.000265,0.00013,0.003167,0.00905,-0.01113,-0.008859,-0.00136,-0.00037,0.001928,0.001796
Turkish_Rumeli:Turkish_Kircaali3,0.10927,0.092413, 0.008297,-0.012597,0.001231,-0.001952,-0.00376,0.004384,0.000614,0.007654,-0.000974,-0.000749,-0.001041,0.003853,-0.003393,0.012994,0.012517,-0.001267,0.002263,0.002376,-0.004118,-0.000124,0.005053,-0.002048,-0.003233
Turkish_Rumeli:Turkish_Kircaali4,0.114961,0.084289 ,0.015085,-0.010013,0.01231,-0.006414,0.006815,0.003,-0.008181,-0.001822,-0.006008,0.006594,-0.004608,0.006468,-0.007193,0.006895,0.01004,-0.009628,0.000126,0.007754,-0.004118,0.002844,0.004683,0.003374,-0.005389
Turkish_Rumeli:Turkish_Gumulcine1,0.100164,0.06600 9,0.012068,-0.008398,-0.010156,-0.010877,0.001175,0.007154,0,0.003827,-0.000325,0.003297,-0.001338,-0.005505,-0.005565,0.00411,0.000391,0.000887,0.006034,0.0002 5,-0.006364,0.000124,0.001725,0.006868,-0.000958
Turkish_Rumeli:Turkish_Gumulcine2,0.103579,0.08632 ,0.013199,-0.010982,-0.002462,-0.009761,0.00094,0.005077,-0.006136,0.000911,0.000162,-0.000749,0.002379,-0.005505,-0.012486,0.013524,0.015255,0.00152,0.007668,0.0043 77,-0.004866,-0.000247,0,0.010001,-0.002155
Turkish_Rumeli:Turkish_Gumulcine3,0.108132,0.08835 1,0.009051,-0.010013,-0.003693,-0.007251,0.001175,0.004384,-0.009204,0.005832,-0.000974,-3e-04,-0.001933,-0.004542,-0.018051,0.020286,0.013169,-0.00038,0.013198,0.005878,0.003244,0.003339,-0.00493,0.007471,0.000838
AVG
Turkish_Rumeli,0.108764111,0.092413222,0.012026,-0.008254444,0.003248222,-0.006476333,0.001853889,0.002820333,-0.004113333,0.002875333,-0.000523222,0.000899111,0.000710222,0.000351444,-0.010073444,0.008353111,0.009851333,-0.000647556,0.005447,0.002223222,-0.005129778,0.001538889,0.001013222,0.002972333,3. 97778E-05

UNSCALED
Turkish_Deliorman:Turkish_Silistre1,0.0096,0.0088, 0.0054,0.0008,0.0033,0.0001,0.0021,0.0001,-0.0037,0.0008,-0.0061,-0.0037,0.0002,0.0109,-0.0067,0,0.0073,0.0022,0.0012,-0.0045,-0.0088,0.0023,-0.0008,-0.0015,-0.001
Turkish_Deliorman:Turkish_Silistre2,0.0103,0.01,0. 0056,0.0019,0.005,0.0015,0.0037,0.0024,-0.0032,-0.0041,-0.0016,-0.0006,0.0055,0.0055,-0.0037,-0.0003,0.0036,0.0008,0.0018,0.0049,-0.0023,0.0027,-0.0002,0.003,0.0009
Turkish_Deliorman:Turkish_Sumnu1,0.0098,0.0104,0.0 032,-0.0017,0.0018,-0.0005,0.0036,0.001,-0.0038,-0.0004,-0.0004,-0.0027,-0.0024,0.0106,-0.0074,-0.0108,-0.0074,0.0014,0.0022,-0.006,-0.0103,-0.0005,-0.0006,-0.0037,-0.0017
Turkish_Deliorman:Turkish_Varna1,0.0096,0.01,0.003 3,-0.0007,0.0002,0.0022,0.0024,0.0035,-0.0027,0.0016,-0.0013,-0.0016,0.001,0.0027,-0.0038,0.0078,0.0098,0.0007,-0.0008,-0.0019,-0.0057,-0.0005,0.0002,0.0062,-0.0062
Turkish_Deliorman:Turkish_Eskicuma1,0.0098,0.0091, 0.0054,0.001,0.0044,0.0013,0.001,0.0014,-0.0015,0.0016,-0.0007,-0.0048,0.0094,-0.0006,-0.0045,0.0026,0.0106,0.0041,0.0029,0.0001,0.0063,-0.0046,-0.0013,0.0026,0.0002
Turkish_Deliorman:Turkish_Razgrad1,0.01,0.0115,0.0 061,-0.0014,0.0032,0.0004,0.0012,0.0051,-0.0013,-0.0032,0.0023,-0.0024,0.0008,0.0054,-0.0076,-0.0093,-0.0064,0.0002,0.0037,-0.0051,-0.0032,-0.003,-0.0047,-0.0004,0.0053
Turkish_Deliorman:Turkish_Razgrad2,0.0096,0.0108,0 .0055,0,0.0065,0.0021,0.0018,0.004,-0.0039,0.0025,0.0018,-0.0015,-0.0004,0.0063,-0.0046,-0.007,-0.0055,0.0011,0.0054,-0.0029,-0.0058,-0.0023,0.0002,0.0023,0.0007
Turkish_Deliorman:Turkish_Razgrad3,0.0096,0.0109,0 .0066,-0.0002,0.0026,-0.0024,0.0018,0.0057,-0.0045,-0.0034,-0.0013,-0.0016,0.0024,0.005,-0.0124,-0.0046,0.0023,0.0013,-0.0015,-0.0002,-0.0015,-0.0016,-0.0003,-0.0032,0.0006
AVG
Turkish_Deliorman,0.0097875,0.0101875,0.0051375,-3.75E-05,0.003375,0.0005875,0.0022,0.0029,-0.003075,-0.000575,-0.0009125,-0.0023625,0.0020625,0.005725,-0.0063375,-0.0027,0.0017875,0.001475,0.0018625,-0.00195,-0.0039125,-0.0009375,-0.0009375,0.0006625,-0.00015

Turkish_Rumeli:Turkish_Selanik1,0.0092,0.0123,0.00 17,-0.003,0.0026,-0.0036,0.0018,-0.0015,-0.001,0.0027,0.0056,0.0023,-0.0007,0.004,-0.0052,-0.0003,0.0017,0.0067,0.0066,-0.0014,-0.0053,0.0024,-0.0013,-0.0035,0.0027
Turkish_Rumeli:Turkish_Kilkis1,0.0104,0.0109,0.002 3,-0.0034,0.0018,0,0.001,0.0019,-0.0007,-0.0001,0.0014,-0.0025,0.0012,0.0044,-0.0135,0.0066,0.0158,-0.0075,-0.0004,0.0024,-0.0077,0.0034,0.0012,0.001,0.0002
Turkish_Rumeli:Turkish_Kırcaali1,0.0093,0.0086,0.0 065,-0.0006,0.0022,-0.0043,-0.0005,0.0008,-0.0026,0.002,-0.0029,0.0004,0.0023,-0.0033,-0.0083,0.0066,0.011,0.0007,0.0023,0.0074,-0.0039,0.0017,0.0025,0.0018,0.005
Turkish_Rumeli:Turkish_Kırcaali2,0.01,0.009,0.0029 ,0.0001,0.0038,0,0.0021,-0.0006,-0.0026,0.0006,-0.002,-0.0002,0.0059,0.001,-0.0054,0.0002,0.0001,0.0025,0.0072,-0.0089,-0.0071,-0.0011,-0.0003,0.0016,0.0015
Turkish_Rumeli:Turkish_Kırcaali3,0.0096,0.0091,0.0 022,-0.0039,0.0004,-0.0007,-0.0016,0.0019,0.0003,0.0042,-0.0006,-0.0005,-0.0007,0.0028,-0.0025,0.0098,0.0096,-0.001,0.0018,0.0019,-0.0033,-0.0001,0.0041,-0.0017,-0.0027
Turkish_Rumeli:Turkish_Kircaali4,0.0101,0.0083,0.0 04,-0.0031,0.004,-0.0023,0.0029,0.0013,-0.004,-0.001,-0.0037,0.0044,-0.0031,0.0047,-0.0053,0.0052,0.0077,-0.0076,0.0001,0.0062,-0.0033,0.0023,0.0038,0.0028,-0.0045
Turkish_Rumeli:Turkish_Gumulcine1,0.0088,0.0065,0. 0032,-0.0026,-0.0033,-0.0039,0.0005,0.0031,0,0.0021,-0.0002,0.0022,-0.0009,-0.004,-0.0041,0.0031,0.0003,0.0007,0.0048,0.0002,-0.0051,0.0001,0.0014,0.0057,-0.0008
Turkish_Rumeli:Turkish_Gumulcine2,0.0091,0.0085,0. 0035,-0.0034,-0.0008,-0.0035,0.0004,0.0022,-0.003,0.0005,0.0001,-0.0005,0.0016,-0.004,-0.0092,0.0102,0.0117,0.0012,0.0061,0.0035,-0.0039,-0.0002,0,0.0083,-0.0018
Turkish_Rumeli:Turkish_Gumulcine3,0.0095,0.0087,0. 0024,-0.0031,-0.0012,-0.0026,0.0005,0.0019,-0.0045,0.0032,-0.0006,-0.0002,-0.0013,-0.0033,-0.0133,0.0153,0.0101,-0.0003,0.0105,0.0047,0.0026,0.0027,-0.004,0.0062,0.0007
AVG
Turkish_Rumeli,0.009555556,0.0091,0.003188889,-0.002555556,0.001055556,-0.002322222,0.000788889,0.001222222,-0.002011111,0.001577778,-0.000322222,0.0006,0.000477778,0.000255556,-0.007422222,0.0063,0.007555556,-0.000511111,0.004333333,0.001777778,-0.004111111,0.001244444,0.000822222,0.002466667,3. 33333E-05


https://i.ibb.co/0K4NjZp/Rumeli-horz.jpg

https://i.ibb.co/54KDYQv/Ads-z.png

P.S: Unscaled coordinates works better for Turks.

Kyp
07-21-2020, 11:21 AM
P.S: Unscaled coordinates works better for Turks.


Why is that?

Kaspias
07-21-2020, 11:52 AM
Why is that?

Scaled coordinates overestimate the difference between East Eurasian and West Eurasian. Despite it might be good while using the ADMIXTURE method it is definitely worse on single population sharing. See what I mean:

Distance to: Turkish_Rumeli - Scaled
0.02739552 Turkish_Deliorman
0.04410071 Turkish_Istanbul
0.05299047 Gagauz
0.05372140 Roma_Barcelona
0.05547328 Turkish_Balikesir
0.05621853 Turkish_Northwest
0.05713367 Bulgarian
0.05819786 Greek_Central_Macedonia
0.05952606 Ashkenazi_Russia
0.06002166 Turkish_Aydin
0.06065077 Turkish_Southwest
0.06091757 Ashkenazi_Belarussia
0.06093748 Albanian
0.06129411 Ashkenazi_Lithuania
0.06161389 Romanian
0.06303030 Greek_Thessaly
0.06366139 Greek_Peloponnese
0.06387945 Ashkenazi_Ukraine
0.06416716 Greek_Izmir
0.06443054 Swiss_Italian
0.06501779 Turkish_North
0.06553068 Roma_Madrid
0.06564026 Macedonian
0.06625456 Moldovan
0.06649431 Turkish_Central

Kyp
07-21-2020, 11:55 AM
Scaled coordinates overestimate the difference between East Eurasian and West Eurasian. Despite it might be good while using the ADMIXTURE method it is definitely worse on single population sharing. See what I mean:

Distance to: Turkish_Rumeli - Scaled
0.02739552 Turkish_Deliorman
0.04410071 Turkish_Istanbul
0.05299047 Gagauz
0.05372140 Roma_Barcelona
0.05547328 Turkish_Balikesir
0.05621853 Turkish_Northwest
0.05713367 Bulgarian
0.05819786 Greek_Central_Macedonia
0.05952606 Ashkenazi_Russia
0.06002166 Turkish_Aydin
0.06065077 Turkish_Southwest
0.06091757 Ashkenazi_Belarussia
0.06093748 Albanian
0.06129411 Ashkenazi_Lithuania
0.06161389 Romanian
0.06303030 Greek_Thessaly
0.06366139 Greek_Peloponnese
0.06387945 Ashkenazi_Ukraine
0.06416716 Greek_Izmir
0.06443054 Swiss_Italian
0.06501779 Turkish_North
0.06553068 Roma_Madrid
0.06564026 Macedonian
0.06625456 Moldovan
0.06649431 Turkish_Central

okay.

This is my father's unscaled:

Distance to: FatherKyp_unscaled
0.01478901 Turkish_Adana
0.01487034 Azeri
0.01686212 Turkish_Kayseri
0.01688834 Ezid
0.01765514 Kurdish
0.01769442 Iranian_Lor
0.01793767 Talysh_Azerbaijan
0.01798871 Turkish_Central
0.01884441 Turkish_Aydin
0.01888389 Iranian_Fars
0.01898451 Turkish_North
0.01942379 Turkish_Istanbul
0.01957037 Turkish_Southwest
0.01993982 Azeri_Turkey
0.02051646 Armenian
0.02056931 Iranian_Zoroastrian
0.02063019 Turkish_South
0.02083587 Assyrian
0.02138816 Iranian_Seyyed
0.02156904 Turkish_Balikesir
0.02178547 Turkish_Northwest
0.02185117 Iranian_Jew
0.02204412 Azeri_Dagestan
0.02290266 Georgian_Jew
0.02326081 Iranian_Mazandarani

Suprised to see Adana as #1. Also closer to Aydin than to Fars despite being Azeri+Persian.

Kaspias
07-21-2020, 12:08 PM
okay.

This is my father's unscaled:

Distance to: FatherKyp_unscaled
0.01478901 Turkish_Adana
0.01487034 Azeri
0.01686212 Turkish_Kayseri
0.01688834 Ezid
0.01765514 Kurdish
0.01769442 Iranian_Lor
0.01793767 Talysh_Azerbaijan
0.01798871 Turkish_Central
0.01884441 Turkish_Aydin
0.01888389 Iranian_Fars
0.01898451 Turkish_North
0.01942379 Turkish_Istanbul
0.01957037 Turkish_Southwest
0.01993982 Azeri_Turkey
0.02051646 Armenian
0.02056931 Iranian_Zoroastrian
0.02063019 Turkish_South
0.02083587 Assyrian
0.02138816 Iranian_Seyyed
0.02156904 Turkish_Balikesir
0.02178547 Turkish_Northwest
0.02185117 Iranian_Jew
0.02204412 Azeri_Dagestan
0.02290266 Georgian_Jew
0.02326081 Iranian_Mazandarani

Suprised to see Adana as #1. Also closer to Aydin than to Fars despite being Azeri+Persian.

Adana sample includes Kurds and Kurdish mixed individuals too. A mix of Turkic and Iranic apparently made it closer to your father.

As I said, PCA method itself takes highly into account the difference between East and West Eurasia which reflects unscaled coordinates too. (Scaled makes it extreme level) If person has a considerable amount of East Eurasian admixture, it will be shifted to other communities who carry a similar admixture no matter how native profile is different. Again, see me:

Distance to: Kaspias
0.01072238 Turkish_Rumeli
0.01973576 Turkish_Southwest
0.02139073 Greek_Izmir
0.02250105 Greek_Peloponnese
0.02251015 Turkish_Northwest
0.02262022 Greek_Central_Macedonia
0.02291710 Turkish_Istanbul
0.02314752 Moldovan
0.02320979 Turkish_Deliorman
0.02451540 Turkish_South

Distance to: Kaspias_scaled
0.01746150 Turkish_Rumeli
0.04075409 Turkish_Deliorman
0.04790357 Turkish_Istanbul
0.05314074 Turkish_Northwest
0.05338578 Turkish_Balikesir
0.05540415 Turkish_Southwest
0.05663089 Turkish_Aydin
0.06336242 Turkish_North
0.06423715 Roma_Barcelona
0.06592998 Turkish_Central


Do you think these oracles make sense for me?

Kyp
07-21-2020, 12:29 PM
Adana sample includes Kurds and Kurdish mixed individuals too. A mix of Turkic and Iranic apparently made it closer to your father.

Okay that makes sense.

Target: Turkish_Adana
Distance: 0.4560% / 0.00455977 | ADC: 0.5x
62.0 Turkish_Kayseri
13.0 Turkish_Central
10.0 Turkish_Southwest
8.6 Iranian_Lor
6.2 Turkish_Istanbul
0.2 Ethiopian_Wolayta



As I said, PCA method itself takes highly into account the difference between East and West Eurasia which reflects unscaled coordinates too. (Scaled makes it extreme level) If person has a considerable amount of East Eurasian admixture, it will be shifted to other communities who carry a similar admixture no matter how native profile is different. Again, see me:

Distance to: Kaspias
0.01072238 Turkish_Rumeli
0.01973576 Turkish_Southwest
0.02139073 Greek_Izmir
0.02250105 Greek_Peloponnese
0.02251015 Turkish_Northwest
0.02262022 Greek_Central_Macedonia
0.02291710 Turkish_Istanbul
0.02314752 Moldovan
0.02320979 Turkish_Deliorman
0.02451540 Turkish_South

Distance to: Kaspias_scaled
0.01746150 Turkish_Rumeli
0.04075409 Turkish_Deliorman
0.04790357 Turkish_Istanbul
0.05314074 Turkish_Northwest
0.05338578 Turkish_Balikesir
0.05540415 Turkish_Southwest
0.05663089 Turkish_Aydin
0.06336242 Turkish_North
0.06423715 Roma_Barcelona
0.06592998 Turkish_Central


Do you think these oracles make sense for me?

At first glance, they make sense to me. Your foreign admixture probably drifts you apart from other Europeans like you said, resulting in a little bit corrupted results maybe?

Rgvgjhvv
07-21-2020, 01:08 PM
Which G25 calculator is being used here?

Hapanuwa
07-21-2020, 01:26 PM
P.S: Unscaled coordinates works better for Turks.


I don't know about my unscaled results. I'm closer to Central Turkey than Western Turkey, even though I'm Egeli. Also it's weird that Italian samples are closer to me than Greek_Izmir

Distance to: Hapanuwa
0.01278771 Turkish_Istanbul
0.01437723 Turkish_Central
0.01447471 Turkish_Northwest
0.01520318 Turkish_Balikesir
0.01544930 Turkish_Aydin
0.01571740 Turkish_North
0.01629638 Turkish_South
0.01705858 Turkish_Kayseri
0.01734694 Turkish_Southwest
0.01742548 Turkish_Adana
0.01977316 Turkish_East
0.01977388 Ashkenazi_Belarussia
0.02002565 Greek_Crete
0.02029520 Ashkenazi_Lithuania
0.02065070 Greek_Central_Anatolia
0.02069085 Greek_Kos
0.02074170 Turkish_Rumeli
0.02104619 Ashkenazi_Russia
0.02145337 Italian_Basilicata
0.02164659 Italian_Abruzzo
0.02170694 Greek_Izmir
0.02172042 Ashkenazi_Poland
0.02173827 Ashkenazi_Ukraine
0.02193982 Italian_Campania
0.02210341 Italian_Apulia

vbnetkhio
07-21-2020, 01:43 PM
I asked an update on Turkish_Balkans samples recently.


why doesn't he update other Balkan populations too?

Kaspias
07-21-2020, 01:56 PM
Okay that makes sense.

Target: Turkish_Adana
Distance: 0.4560% / 0.00455977 | ADC: 0.5x
62.0 Turkish_Kayseri
13.0 Turkish_Central
10.0 Turkish_Southwest
8.6 Iranian_Lor
6.2 Turkish_Istanbul
0.2 Ethiopian_Wolayta



At first glance, they make sense to me. Your foreign admixture probably drifts you apart from other Europeans like you said, resulting in a little bit corrupted results maybe?

They does make sense when you know the glitch here. The same goes for your father; that's why he got Aydın before Persian.





Which G25 calculator is being used here?

No specific calculator used. Just single pop run by using all populations from the official source: https://eurogenes.blogspot.com/2019/07/getting-most-out-of-global25_12.html



I don't know about my unscaled results. I'm closer to Central Turkey than Western Turkey, even though I'm Egeli. Also it's weird that Italian samples are closer to me than Greek_Izmir

Distance to: Hapanuwa
0.01278771 Turkish_Istanbul
0.01437723 Turkish_Central
0.01447471 Turkish_Northwest
0.01520318 Turkish_Balikesir
0.01544930 Turkish_Aydin
0.01571740 Turkish_North
0.01629638 Turkish_South
0.01705858 Turkish_Kayseri
0.01734694 Turkish_Southwest
0.01742548 Turkish_Adana
0.01977316 Turkish_East
0.01977388 Ashkenazi_Belarussia
0.02002565 Greek_Crete
0.02029520 Ashkenazi_Lithuania
0.02065070 Greek_Central_Anatolia
0.02069085 Greek_Kos
0.02074170 Turkish_Rumeli
0.02104619 Ashkenazi_Russia
0.02145337 Italian_Basilicata
0.02164659 Italian_Abruzzo
0.02170694 Greek_Izmir
0.02172042 Ashkenazi_Poland
0.02173827 Ashkenazi_Ukraine
0.02193982 Italian_Campania
0.02210341 Italian_Apulia

Can you run unscaled again by using this spreadsheet: https://drive.google.com/file/d/1y49hyvviJpHj9esVqyeiFm32DhnPlfRQ/view

Also, what about scaled results?

Greek_Izmir samples are not actual Greeks from Izmir, but recent comers from the Mainland.



why doesn't he update other Balkan populations too?

I'm supplying raw data myself, he just runs and copies into spreadsheet. If you have raw data I can ask him to do.

Rgvgjhvv
07-21-2020, 02:00 PM
Greek_Izmir samples fill the "gap" between Mainlanders and Islanders and therefore are most likely a mix of both.

Ion Basescul
07-21-2020, 02:04 PM
P.S: Unscaled coordinates works better for Turks.


But what makes you certain of unscaled coords being better? I'm asking since everyone has different pops at the top with unscaled coords. For me personally, scaled coordinates are better, as they fall in-line with the pops that I get at the top on GEDmatch calculators. Unscaled coords make me too North-shifted.

Distance to: Ion_Basescul_23andme+LivingDNA
0.03076421 Moldovan
0.03236514 Montenegrin
0.03292518 Bosnian
0.03711987 Romanian
0.03849710 Croatian
0.03899393 Hungarian
0.03925396 Serbian
0.04152364 Austrian
0.04205013 Slovenian
0.04313328 Bulgarian

Distance to: Ion_Basescul_23andme+LivingDNA_unscaled
0.01444933 Montenegrin
0.01716390 Bosnian
0.01755487 Slovakian
0.01878613 Romanian
0.01970101 Macedonian
0.01999210 Moldovan
0.02025696 Bulgarian
0.02041216 Moldovan_o_UkrainianMixes
0.02078243 Serbian
0.02144482 Croatian

Kaspias
07-21-2020, 02:13 PM
But what makes you certain of unscaled coords being better? I'm asking since everyone has different pops at the top with unscaled coords. For me personally, scaled coordinates are better, as they fall in-line with the pops that I get at the top on GEDmatch calculators. Unscaled coords make me too North-shifted.

Distance to: Ion_Basescul_23andme+LivingDNA
0.03076421 Moldovan
0.03236514 Montenegrin
0.03292518 Bosnian
0.03711987 Romanian
0.03849710 Croatian
0.03899393 Hungarian
0.03925396 Serbian
0.04152364 Austrian
0.04205013 Slovenian
0.04313328 Bulgarian

Distance to: Ion_Basescul_23andme+LivingDNA_unscaled
0.01444933 Montenegrin
0.01716390 Bosnian
0.01755487 Slovakian
0.01878613 Romanian
0.01970101 Macedonian
0.01999210 Moldovan
0.02025696 Bulgarian
0.02041216 Moldovan_o_UkrainianMixes
0.02078243 Serbian
0.02144482 Croatian


P.S: Unscaled coordinates works better for Turks.


Scaled coordinates work better while using admixture breakdown for everyone; scaled coordinates might work better in single population sharing of Europeans as they are only West Eurasian; unscaled coordinates work better for the people who carry both East Eurasian and West Eurasian admixture together as eigenvalue option tend to overestimate the difference between populations.

I think I already posted enough examples of why I believe it so.

Rgvgjhvv
07-21-2020, 02:14 PM
Distance to: Markos_Mother
0.01847539 Turkish_Deliorman:Turkish_Razgrad2
0.01862720 Greek_Thessaly
0.01926932 Sicilian_East
0.01966440 Gagauz
0.02046973 Turkish_Deliorman:Turkish_Razgrad1
0.02088048 Ashkenazi_Ukraine
0.02097556 Italian_Apulia
0.02115372 Turkish_Deliorman:Turkish_Sumnu1
0.02120330 Turkish_Rumeli:Turkish_Selanik1
0.02127122 Albanian
0.02128850 Sicilian_West
0.02167210 Italian_Liguria
0.02194079 Ashkenazi_Poland
0.02220839 Greek_Crete
0.02227453 Italian_Marche
0.02244769 Greek_Central_Macedonia
0.02272459 Italian_Abruzzo
0.02293888 Bulgarian
0.02347001 Italian_Basilicata
0.02362315 Maltese
0.02383287 Italian_Campania
0.02392655 Italian_Molise
0.02397891 Ashkenazi_Russia
0.02435996 Ashkenazi_Belarussia

Target: Markos_Mother
Distance: 1.6283% / 0.01628312
50.8 Turkish_Deliorman
49.2 Greek_Thessaly

Target: Markos_Mother
Distance: 1.4807% / 0.01480718 | ADC: 2x
57.6 Turkish_Deliorman
30.0 Sicilian_East
12.4 Greek_Thessaly

vbnetkhio
07-21-2020, 02:16 PM
I'm supplying raw data myself, he just runs and copies into spreadsheet. If you have raw data I can ask him to do.

I mean from the academic samples. I know of a dozen samples from Montenegro, Serbia, Bulgaria, Bosnia and Macedonia (he has only 5 of each in the sheet), and he probably has access to even more samples i don't.

He could run them all, and if they can't all be in the sheet, than at least keep the most representative ones. a big part of his users are from the Balkans.

altaic
07-21-2020, 03:01 PM
I tried g25 Unscaled and added 2 more sample (Deliorman and Rumeli). This is what I am getting. Unscaled seems good for me.

G25 Population Modern - Unscaled

Target: Altaic
Distance: 0.3223% / 0.00322320 | ADC: 0.25x
18.2 Turkish_Adana
15.2 Turkish_Kayseri
13.6 Turkish_Istanbul
13.0 Azeri
12.4 Armenian_Hemsheni
12.2 Greek_Kos
5.2 Roma_Balkans
3.0 Turkish_Southwest
2.2 Iranian_Jew
2.2 Iranian_Lor
1.4 Chuvash
0.8 Udmurt
0.4 Nenets
0.2 Mari

Target: Altaic
Distance: 0.5392% / 0.00539155 | ADC: 0.5x
26.2 Turkish_Kayseri
14.4 Turkish_Istanbul
9.0 Azeri
8.6 Turkish_North
8.2 Turkish_Adana
7.6 Turkish_Central
6.0 Turkish_Southwest
5.8 Ezid
5.2 Kurdish
5.0 Iranian_Jew
3.2 Greek_Kos
0.4 Iranian_Lor
0.4 Turkish_East

-------------------------------------------------

G25 Population Averages Modern - Unscaled

Target: Altaic
Distance: 0.6328% / 0.00632816 | ADC: 0.25x
28.2 Greek_Central_Anatolia
22.0 Turkish_East
18.0 Turkish_Adana
10.2 Armenian_Hemsheni
8.6 Makrani
4.8 Georgian_Jew
4.2 Chuvash
1.6 Ezid
1.2 Aeta
1.0 Cossack_Kuban
0.2 Surui

Target: Altaic
Distance: 0.8951% / 0.00895109 | ADC: 0.5x
42.6 Turkish_East
37.2 Turkish_Adana
11.6 Greek_Central_Anatolia
7.6 Ezid
1.0 Chuvash

Target: Altaic
Distance: 0.9685% / 0.00968457 | ADC: 1x
50.8 Turkish_East
49.2 Turkish_Adana

Ion Basescul
07-21-2020, 03:02 PM
Scaled coordinates work better while using admixture breakdown for everyone; scaled coordinates might work better in single population sharing of Europeans as they are only West Eurasian; unscaled coordinates work better for the people who carry both East Eurasian and West Eurasian admixture together as eigenvalue option tend to overestimate the difference between populations.

I think I already posted enough examples of why I believe it so.

Is this a joke? Gagauz people have less East Eurasian ancestry than even Moldovans/Romanians. And that can be seen anywhere from Global 25 to GEDmatch and academic studies. Turks from Deliorman match them and other Southeast Europeans before other Turks on unscaled.

Turkish Deliorman logically pulls to Turkish Rumeli on scaled, since they are very similar.


Distance to: Turkish_Deliorman
0.02739549 Turkish_Rumeli
0.03590543 Gagauz
0.04084561 Bulgarian
0.04438533 Romanian
0.04522681 Moldovan
0.04846759 Macedonian
0.05001280 Serbian
0.05031737 Greek_Central_Macedonia
0.05187104 Montenegrin
0.05295266 Albanian


https://i.ibb.co/L0MSDnk/image.png

On unscaled, Turks from Deliorman are closer to Southeast Europeans than to Turks from Rumeli. It is exactly how it works for me in my example. Unscaled coords pulls them further to the North than where they should be.


Distance to: Turkish_Deliorman
0.00944847 Gagauz
0.01150207 Bulgarian
0.01270549 Romanian
0.01309212 Moldovan
0.01397265 Macedonian
0.01398290 Serbian
0.01509710 Turkish_Rumeli
0.01595943 Montenegrin
0.01624582 Greek_Central_Macedonia
0.01817576 Bosnian

Kaspias
07-21-2020, 03:20 PM
Is this a joke? Gagauz people have less East Eurasian ancestry than even Moldovans/Romanians. And that can be seen anywhere from Global 25 to GEDmatch and academic studies. Turks from Deliorman match them and other Southeast Europeans before other Turks on unscaled.

Turkish Deliorman logically pulls to Turkish Rumeli on scaled, since they are very similar.


Distance to: Turkish_Deliorman
0.02739549 Turkish_Rumeli
0.03590543 Gagauz
0.04084561 Bulgarian
0.04438533 Romanian
0.04522681 Moldovan
0.04846759 Macedonian
0.05001280 Serbian
0.05031737 Greek_Central_Macedonia
0.05187104 Montenegrin
0.05295266 Albanian


https://i.ibb.co/L0MSDnk/image.png

On unscaled, Turks from Deliorman are closer to Southeast Europeans than to Turks from Rumeli. It is exactly how it works for me in my example. Unscaled coords pulls them further to the North than where they should be.


Distance to: Turkish_Deliorman
0.00944847 Gagauz
0.01150207 Bulgarian
0.01270549 Romanian
0.01309212 Moldovan
0.01397265 Macedonian
0.01398290 Serbian
0.01509710 Turkish_Rumeli
0.01595943 Montenegrin
0.01624582 Greek_Central_Macedonia
0.01817576 Bosnian

I don't understand what are you talking about. What is the relevance with Gagauz?

Deliorman Turks should be closer to Gagauz because both are from the same region while Rumelian Turks includes a bunch of Greek admixture. They are not very similar. Deliorman gets Rumeli in scaled because of that PCA effect I stated.

Maybe you should re-read what I wrote.

Rgvgjhvv
07-21-2020, 03:22 PM
Distance to: Markos_Mother
0.01847539 Turkish_Deliorman:Turkish_Razgrad2
0.01862720 Greek_Thessaly
0.01926932 Sicilian_East
0.01966440 Gagauz
0.02046973 Turkish_Deliorman:Turkish_Razgrad1
0.02088048 Ashkenazi_Ukraine
0.02097556 Italian_Apulia
0.02115372 Turkish_Deliorman:Turkish_Sumnu1
0.02120330 Turkish_Rumeli:Turkish_Selanik1
0.02127122 Albanian
0.02128850 Sicilian_West
0.02167210 Italian_Liguria
0.02194079 Ashkenazi_Poland
0.02220839 Greek_Crete
0.02227453 Italian_Marche
0.02244769 Greek_Central_Macedonia
0.02272459 Italian_Abruzzo
0.02293888 Bulgarian
0.02347001 Italian_Basilicata
0.02362315 Maltese
0.02383287 Italian_Campania
0.02392655 Italian_Molise
0.02397891 Ashkenazi_Russia
0.02435996 Ashkenazi_Belarussia

Target: Markos_Mother
Distance: 1.6283% / 0.01628312
50.8 Turkish_Deliorman
49.2 Greek_Thessaly


The distance becomes slightly closer when I add Sicilian_East instead, however:

Target: Markos_Mother
Distance: 1.4457% / 0.01445688
52.6 Turkish_Deliorman
47.4 Sicilian_East

From this, I would assume the Turkish_Deliorman is being linked to her moreso for its Balkan component, so instead, I try and use a Bulgarian sample to see if the fit stays the same.

Target: Markos_Mother
Distance: 1.8502% / 0.01850240
86.2 Greek_Thessaly
13.8 Bulgarian

Not even close!

With Gagauz instead of Bulgarian:

Target: Markos_Mother
Distance: 1.7643% / 0.01764332
59.2 Greek_Thessaly
40.8 Gagauz

Much better than the Bulgarian fit, but the Turkish_Deliorman increases and betters the fit by 10%. So perhaps the conclusion is the minor EE component and similarity in the Balkan Turk sample is bringing her results closer to it.

Given my grandfather's village's history of mixed Greek/Turk/Albanian(?) marriages are concerned, it's not out of the question to think their mixing would have given them a similar profile to modern-day Balkan Turks. In that sense, my mother could realistically be 'described' as 50% Greek Proper and 50% Balkan Turk. Is this a stretch?

Impaler
07-21-2020, 03:28 PM
Distance to: Impaler
0.01443955 Romanian
0.01504495 Bulgarian
0.01543567 Turkish_Deliorman:Turkish_Varna1
0.01545607 Turkish_Rumeli:Turkish_Kırcaali2
0.01577910 Turkish_Rumeli:Turkish_Gumulcine1
0.01631980 Italian_Northeast
0.01674449 Greek_Central_Macedonia
0.01700978 Gagauz
0.01716150 Montenegrin
0.01725370 Turkish_Deliorman:Turkish_Silistre2
0.01761566 Macedonian
0.01767371 Turkish_Rumeli:Turkish_Kırcaali3
0.01802818 Serbian
0.01812298 Greek_Thessaly
0.01865878 Turkish_Deliorman:Turkish_Razgrad2
0.01887829 Turkish_Deliorman:Turkish_Silistre1
0.01912439 Swiss_Italian
0.01930104 Turkish_Rumeli:Turkish_Selanik1
0.01931729 Moldovan
0.01938401 Turkish_Deliorman:Turkish_Eskicuma1
0.01939510 Turkish_Rumeli:Turkish_Kırcaali1
0.01946160 Albanian
0.02005507 Italian_Veneto
0.02030545 Greek_Izmir
0.02035682 Turkish_Rumeli:Turkish_Gumulcine2

Ion Basescul
07-21-2020, 03:42 PM
I don't understand what are you talking about. What is the relevance with Gagauz?

Deliorman Turks should be closer to Gagauz because both are from the same region while Rumelian Turks includes a bunch of Greek admixture. They are not very similar. Deliorman gets Rumeli in scaled because of that PCA effect I stated.

Maybe you should re-read what I wrote.

The relevance is in Deliorman and Rumeli Turks being closer to each other than Deliorman Turks being to Gagauz, Bulgarians, Romanians, Moldovans, Macedonians and Serbs.
I claimed that unscaled coords pull me to the North, which is different from all other tests. You claimed that in the case of an West and East Eurasian admixed population, unscaled coords work better.
Well, they don't. Turks are also pulled artificially to the North. At least that's the case for Deliorman and Rumeli folk, since I haven't checked all the averages.

Kaspias
07-21-2020, 03:49 PM
The relevance is in Deliorman and Rumeli Turks being closer to each other than Deliorman Turks being to Gagauz, Bulgarians, Romanians, Moldovans, Macedonians and Serbs.
I claimed that unscaled coords pull me to the North, which is different from all other tests. You claimed that in the case of an West and East Eurasian admixed population, unscaled coords work better.
Well, they don't. Turks are also pulled artificially to the North. At least that's the case for Deliorman and Rumeli folk, since I haven't checked all the averages.

There are Deliorman samples who are more European than both Bulgarian and Gagauz average. They are getting what they should get in unscaled. Same goes for Rumelia.

Better to check samples before claim something.

Hapanuwa
07-21-2020, 04:07 PM
Can you run unscaled again by using this spreadsheet: https://drive.google.com/file/d/1y49hyvviJpHj9esVqyeiFm32DhnPlfRQ/view

Also, what about scaled results?

Greek_Izmir samples are not actual Greeks from Izmir, but recent comers from the Mainland.

I'm supplying raw data myself, he just runs and copies into spreadsheet. If you have raw data I can ask him to do.

When I run unscaled using spreadsheet I get exactly the same results as the one in my previous post.

My scaled results are as the following:

Distance to: Hapanuwa_scaled
0.02957897 Turkish_North
0.02998521 Turkish_Central
0.03074857 Turkish_Istanbul
0.03099863 Turkish_Aydin
0.03200715 Turkish_Balikesir
0.03259877 Turkish_Northwest
0.03457687 Turkish_Southwest
0.03858371 Turkish_Adana
0.03908243 Turkish_South
0.04022540 Turkish_Kayseri
0.05293626 Turkish_Rumeli
0.05542837 Azeri
0.05779093 Azeri_Turkey
0.05872370 Turkish_East
0.05883086 Kumyk
0.05898158 Kabardin
0.06145902 Azeri_Dagestan
0.06509284 Abazin
0.06514874 Cherkes
0.06530014 Karachay
0.06560161 Circassian
0.06685454 Ashkenazi_Lithuania
0.06692698 Ashkenazi_Russia
0.06813385 Ashkenazi_Belarussia
0.06916516 Turkish_Deliorman

Very weird that Turkish north, central and İstanbul are at the top

Kaspias
07-21-2020, 04:30 PM
I mean from the academic samples. I know of a dozen samples from Montenegro, Serbia, Bulgaria, Bosnia and Macedonia (he has only 5 of each in the sheet), and he probably has access to even more samples i don't.

He could run them all, and if they can't all be in the sheet, than at least keep the most representative ones. a big part of his users are from the Balkans.

I can contact with him if you have enough amount of academic samples which you can either create new components or just add more samples to averages. Just send me the details.

Ion Basescul
07-21-2020, 04:33 PM
There are Deliorman samples who are more European than both Bulgarian and Gagauz average. They are getting what they should get in unscaled. Same goes for Rumelia.

Better to check samples before claim something.

Nope and I already showed that, so I'm not the one who needs to prove otherwise. You are just pushing an agenda here.

Kaspias
07-21-2020, 04:39 PM
Nope and I already showed that, so I'm not the one who needs to prove otherwise. You are just pushing an agenda here.

This is getting irritating.


The relevance is in Deliorman and Rumeli Turks being closer to each other than Deliorman Turks being to Gagauz, Bulgarians, Romanians, Moldovans, Macedonians and Serbs.


This is already what it should be.

Gedmatch is down now, so I'm posting what I'm trying to say by using the samples I provided to Vahaduo K13.

Distance to: Turk_Razgrad
4.50211062 Romania_Muntenia
5.20874265 Romania_Dobruja
5.53974729 Bulgarian_Gabrovo
5.61178225 Romanian_Wallachia
5.64796424 Bulgarian
5.94708332 Bulgarian_North-Rhodope
6.00432344 Moldova_South
6.15556659 Bulgarian_Stara-Zagora
6.23676198 Bulgarian_Eastern-Rhodopes
6.42479572 Romania_Banat
6.58817881 Romanian_Oltenia
6.65087964 Bulgarian_Burgas
6.67506275 Macedonian
6.73691324 Macedonian_Northwest
7.02683428 Macedonian_Southeast
7.24116703 Romania_Transylvania
7.39091334 Bulgarian_Central
7.84624751 Romania_Moldavia_average
7.97998747 Bulgarian_Plovdiv
7.99408531 Romania_Moldavia_South
8.01159160 Romania_Moldavia_North
8.12362604 Bulgarian_Southwest
8.22965977 Macedonian_Central
8.29630038 Bulgarian_Sofia
8.33730772 Serb_Serbia_South&East
8.44353007 Bulgarian_Pleven
8.44515838 Macedonian_East
8.83572295 Macedonian_Greece
8.84407711 Bulgarian_Serres
9.02170161 Moldova_Centre
9.06846183 Macedonian_North
9.11749417 Turk_Thessaloniki
9.25642480 Bulgarian_Pazardzhik
9.31761772 Serb_Herzegovina
9.56221209 Moldova_average
9.65545113 Bulgarian_Thrace
9.91228026 Serb_Serbia
10.02539276 Romania_Crisana
10.27818564 Serb
10.41941457 Bulgarian_Northwest
10.53299578 Turk_Komotini
10.55266791 Bulgarian_Ruse
10.76527752 Macedonian_Southwest
10.79894439 Serb_Serbia_West&Central
10.82846249 Serb_Serbia_Vojvodina
10.90605795 Romania_Maramures
10.91519583 Serb_Croatia_Lika&Dalmatia
10.97489408 Serb_B&H_Krajina
11.03132358 Serb_Croatia
11.05387715 Bulgarian_Dupnitsa
11.14627741 Turk_Kardzhali-Mountaineer
11.49489452 Serb_Croatia_Kordun&Banija
11.72504584 Turk_Kardzhali-South-Plain
12.11456974 Serb_B&H
12.65071144 Serb_B&H_Northeast&Central
14.80109793 Moldova_North


Was that hard to understand it? Or are you trying to troll me?

Kaspias
07-21-2020, 04:59 PM
The distance becomes slightly closer when I add Sicilian_East instead, however:

Target: Markos_Mother
Distance: 1.4457% / 0.01445688
52.6 Turkish_Deliorman
47.4 Sicilian_East

From this, I would assume the Turkish_Deliorman is being linked to her moreso for its Balkan component, so instead, I try and use a Bulgarian sample to see if the fit stays the same.

Target: Markos_Mother
Distance: 1.8502% / 0.01850240
86.2 Greek_Thessaly
13.8 Bulgarian

Not even close!

With Gagauz instead of Bulgarian:

Target: Markos_Mother
Distance: 1.7643% / 0.01764332
59.2 Greek_Thessaly
40.8 Gagauz

Much better than the Bulgarian fit, but the Turkish_Deliorman increases and betters the fit by 10%. So perhaps the conclusion is the minor EE component and similarity in the Balkan Turk sample is bringing her results closer to it.

Given my grandfather's village's history of mixed Greek/Turk/Albanian(?) marriages are concerned, it's not out of the question to think their mixing would have given them a similar profile to modern-day Balkan Turks. In that sense, my mother could realistically be 'described' as 50% Greek Proper and 50% Balkan Turk. Is this a stretch?

This is a good finding, but I couldn't explain how Razgrad1 can pop up first while there is Thessaloniki.

Razgrad1: (one of the samples who supposed to be similar to Thrace and Macedonia according to Ion, and a tool for my secret agenda)

Component %
North_Atlantic 17,96%
Baltic 24,00%
West_Med 13,09%
West_Asian 15,85%
East_Med 20,15%
Red_Sea 2,49%
South_Asian 0,50%
East_Asian 0,05%
Siberian 4,63%
Northeast_African 1,29%

I realized there is also Razgrad2.

Razgrad2, another agenda stuff.

Component %
North_Atlantic 21,22%
Baltic 22,84%
West_Med 14,09%
West_Asian 12,95%
East_Med 19,31%
Red_Sea 1,04%
South_Asian 2,25%
East_Asian 2,56%
Siberian 2,24%
Amerindian 0,25%
Oceanian 0,13%
Northeast_African 0,09%
Sub-Saharan 1,03%


50% of it makes around 2.5% East Eurasian, which corrects your theory. This is full individual run:

https://i.ibb.co/qmqRvc0/markos-unscaled.png

Kaspias
07-21-2020, 05:02 PM
When I run unscaled using spreadsheet I get exactly the same results as the one in my previous post.

My scaled results are as the following:

Distance to: Hapanuwa_scaled
0.02957897 Turkish_North
0.02998521 Turkish_Central
0.03074857 Turkish_Istanbul
0.03099863 Turkish_Aydin
0.03200715 Turkish_Balikesir
0.03259877 Turkish_Northwest
0.03457687 Turkish_Southwest
0.03858371 Turkish_Adana
0.03908243 Turkish_South
0.04022540 Turkish_Kayseri
0.05293626 Turkish_Rumeli
0.05542837 Azeri
0.05779093 Azeri_Turkey
0.05872370 Turkish_East
0.05883086 Kumyk
0.05898158 Kabardin
0.06145902 Azeri_Dagestan
0.06509284 Abazin
0.06514874 Cherkes
0.06530014 Karachay
0.06560161 Circassian
0.06685454 Ashkenazi_Lithuania
0.06692698 Ashkenazi_Russia
0.06813385 Ashkenazi_Belarussia
0.06916516 Turkish_Deliorman

Very weird that Turkish north, central and İstanbul are at the top

Well yeah, it is really weird that you got Istanbul and Central at first. Central is actually weirder than Istanbul, because I could explain it with your suspected Albanian ancestry while I have no explanation for Central Anatolia.

Unscaled looks better though. There is simply no way an Aegean Turk can get Cherkes, Abazin, Kumyk...

Distance to: Hapanuwa
0.01278771 Turkish_Istanbul
0.01437723 Turkish_Central
0.01447471 Turkish_Northwest
0.01520318 Turkish_Balikesir
0.01544930 Turkish_Aydin
0.01571740 Turkish_North
0.01629638 Turkish_South
0.01705858 Turkish_Kayseri
0.01734694 Turkish_Southwest
0.01742548 Turkish_Adana
0.01977316 Turkish_East
0.01977388 Ashkenazi_Belarussia
0.02002565 Greek_Crete
0.02029520 Ashkenazi_Lithuania
0.02065070 Greek_Central_Anatolia
0.02069085 Greek_Kos
0.02074170 Turkish_Rumeli
0.02104619 Ashkenazi_Russia
0.02145337 Italian_Basilicata
0.02164659 Italian_Abruzzo
0.02170694 Greek_Izmir
0.02172042 Ashkenazi_Poland
0.02173827 Ashkenazi_Ukraine
0.02193982 Italian_Campania
0.02210341 Italian_Apulia

vbnetkhio
07-21-2020, 05:07 PM
I can contact with him if you have enough amount of academic samples which you can either create new components or just add more samples to averages. Just send me the details.

ok. i'll wait for gedmatch to get back online so i can check some stuff, then i'll prepare a detailed explanation.

Rgvgjhvv
07-21-2020, 05:09 PM
This is a good finding, but I couldn't explain how Razgrad1 can pop up first while there is Thessaloniki.


It's Razgrad2 which is her first population, unscaled.

Kaspias
07-21-2020, 05:09 PM
It's Razgrad2 which is her first population, unscaled.

I edited.

Rgvgjhvv
07-21-2020, 05:15 PM
I edited.

Oh, okay!

I'm kind of confused about the run you did afterwards. What exactly does it show and is it different to my initial run using G25 modern pop average references? Does it disprove or prove that initial run and breakdown?

Kaspias
07-21-2020, 05:18 PM
Oh, okay!

I'm kind of confused about the run you did afterwards. What exactly does it show and is it different to my initial run using G25 modern pop average references? Does it disprove or prove that initial run and breakdown?

I thought if you want to see the efficiency of individual populations you should run them with other individual populations instead of average. I believe everything is okay regarding her results considering where is she from, but she indeed seems to have 2-2.5% East Eurasian. (My reference here is the distance of Gagauz - Razgrad run)

Don't forget that Greek Peloponnese samples in G25 are not actually from Peloponnese. She picks various other populations because of this reason.

Rgvgjhvv
07-21-2020, 05:26 PM
I thought if you want to see the efficiency of individual populations you should run them with other individual populations instead of average. I believe everything is okay regarding her results considering where is she from, but she indeed seems to have 2-2.5% East Eurasian. (My reference here is the distance of Gagauz - Razgrad run)

Don't forget that Greek Peloponnese samples in G25 are not actually from Peloponnese. She picks various other populations because of this reason.

Ohhh okay yeah I wasn't sure what would be more "accurate" the individual, or the averages. I wonder if her paternal side would come out close to the Razgrad average. Her Ancient Anatolian/W. Asian component is low while still having EE so maybe that is in-line with Balkan Turks? It's hard to say but that 50/50 run was a good fit for her.

Zoro
07-21-2020, 11:45 PM
P.S: Unscaled coordinates works better for Turks.



Well the problem is using unscaled will give you wrong results. It's not permissible to use unscaled in PCA. I realize that so many people are using unscaled here without understanding the basic concepts behind PCA. So basically more than half of the stuff posted here is plain wrong because of it. I simply haven't said anything so far because I don't want to spoil the fun they're having.


Here's why :

So the G25 coordinates are basically PCA coordinates and PCA is a tool used to reduce dimensionality of data to better visualize it.

The actual Plink software Davidski uses does infact output the eigenvalues (scalars) for the eigenvectors.

I'll simplify for easier understanding. Let's assume we have a G3 PCA in other words only 3 coordinates meaning 3 dimensional space.

Let's say you have Ulchi, Turks and Iberians as some of the input populations. So you can plot the position of Iberians, Turks and Ulchi in a 3 dimensional cube. What determines their position in the cube is an Eigenvector (x, y, z) and and Eigenvalue (scalar). For example Iberian positon could be 6x, 4y, 2z. Turks could be 4x, 2y, 3z and so on.

So if you wanted to calculate distance between Iberians and Turks you can do that based on their position in 3D space. So you see Eigenvalues (scalars) are totally essential to do so.

If you take away the eigenvalues and use unscaled then your positions and distances within the cube get screwed up.

That's why you have to use scaled. Now if using scaled you're getting unreasonable results it's because Davidski just willy nilly throws a bunch of samples into the run without regards to sample sizes and other considerations. So using unscaled in this case doesn't compensate for the inherent issues with the run.

Mingle
07-22-2020, 05:16 AM
Scaled coordinates work better while using admixture breakdown for everyone; scaled coordinates might work better in single population sharing of Europeans as they are only West Eurasian; unscaled coordinates work better for the people who carry both East Eurasian and West Eurasian admixture together as eigenvalue option tend to overestimate the difference between populations.

I think I already posted enough examples of why I believe it so.

I'm around somewhere around 15% East Eurasian (mostly ASI with a small amount of Mongoloid) but scaled is much better for me.


Distance to: Mingle_scaled
0.03150615 Pashtun_Uthmankhel
0.03411736 Kalash
0.03544066 Pashtun_Tarkalani
0.03575387 Pashtun_Yusufzai
0.03680886 Pashtun
0.04267163 Kamboj
0.04729382 Kho_Singanali
0.05207452 Ror
0.05465078 Jatt_Pathak
0.05470547 Balochi
0.05492211 Brahui
0.05608177 Punjabi_Jatt
0.05674197 Tajik_Ishkashim
0.05728015 Khatri
0.05847219 Kohistani
0.06227557 Sindhi
0.06648307 Gujar_Pakistan
0.06973125 Kashmiri_Pandit
0.07117605 Parsi_India
0.07151652 Tajik_Shugnan
0.07198459 Parsi_Pakistan
0.07414114 Makrani
0.07691840 Tajik_Rushan
0.07894101 Iranian_Bandari
0.08024164 Tajik


Distance to: Mingle_unscaled
0.01827060 Kalash
0.02066177 Pashtun_Uthmankhel
0.02095253 Pashtun
0.02229768 Kamboj
0.02252060 Kho_Singanali
0.02264234 Pashtun_Yusufzai
0.02320975 Kohistani
0.02437327 Burusho
0.02445550 Ror
0.02475772 Pashtun_Tarkalani
0.02562041 Tajik_Ishkashim
0.02588315 Punjabi_Jatt
0.02632436 Khatri
0.02635375 Jatt_Pathak
0.02815102 Kashmiri_Pandit
0.02845801 Tajik
0.02858206 Tajik_Rushan
0.02866709 Sindhi
0.02907619 Gujar_Pakistan
0.02912045 Tajik_Shugnan
0.02955681 Parsi_Pakistan
0.02988522 Parsi_India
0.03084267 Gujar_India
0.03150619 Tajik_Yagnobi
0.03157725 Iranian_Bandari

For scaled, all the Pashtun samples are ranked close to one another at the top while for unscaled, there's less consistency. The scaled also matches better with the results on GEDmatch. I've yet to see samples where unscaled was better than scaled. I don't think there's any overestimation in the distances for scaled, them having extreme differences makes sense.

Kaspias
07-22-2020, 11:08 AM
Ohhh okay yeah I wasn't sure what would be more "accurate" the individual, or the averages. I wonder if her paternal side would come out close to the Razgrad average. Her Ancient Anatolian/W. Asian component is low while still having EE so maybe that is in-line with Balkan Turks? It's hard to say but that 50/50 run was a good fit for her.

It should be just coincidence that she got Deliorman. She models like 50% Sicilian(or any other population who are East Med dominant) + 50% Gagauz-Balkan Turk(have a mainlander base) which makes sense according where she is from. But having additional East Eurasian make distance better while modeling with Balkan Turks. Should this mean she has a legit Balkan Turk ancestor? I don't think so, because Turkic ancestor supposedly migrated from Anatolia to there. But I can confirm East Eurasian.


Well the problem is using unscaled will give you wrong results. It's not permissible to use unscaled in PCA. I realize that so many people are using unscaled here without understanding the basic concepts behind PCA. So basically more than half of the stuff posted here is plain wrong because of it. I simply haven't said anything so far because I don't want to spoil the fun they're having.


Here's why :

So the G25 coordinates are basically PCA coordinates and PCA is a tool used to reduce dimensionality of data to better visualize it.

The actual Plink software Davidski uses does infact output the eigenvalues (scalars) for the eigenvectors.

I'll simplify for easier understanding. Let's assume we have a G3 PCA in other words only 3 coordinates meaning 3 dimensional space.

Let's say you have Ulchi, Turks and Iberians as some of the input populations. So you can plot the position of Iberians, Turks and Ulchi in a 3 dimensional cube. What determines their position in the cube is an Eigenvector (x, y, z) and and Eigenvalue (scalar). For example Iberian positon could be 6x, 4y, 2z. Turks could be 4x, 2y, 3z and so on.

So if you wanted to calculate distance between Iberians and Turks you can do that based on their position in 3D space. So you see Eigenvalues (scalars) are totally essential to do so.

If you take away the eigenvalues and use unscaled then your positions and distances within the cube get screwed up.

That's why you have to use scaled. Now if using scaled you're getting unreasonable results it's because Davidski just willy nilly throws a bunch of samples into the run without regards to sample sizes and other considerations. So using unscaled in this case doesn't compensate for the inherent issues with the run.

You don't need to explain this to me, because I already know, and actually that's why I recommend unscaled coordinates for Turks. (I have said Turks at the beginning, but everybody taken it except for Turks for some reason?)

The variance for instance in the first global dimension, that is Africa, is set to 1, and the same is set for the second global dimension, Western Eurasia, the same goes for the other, Eastern Eurasia. Scaled definitely makes distances more realistic, and works better among a group(dimension), but the populations such as Tatars and Turks who plot in between two dimensions will be influenced in negative meaning if we apply eigenvalue. How so? They get closer to other "bridge" populations no matter how their base genetic profile is.

Let me show you:


Distance to: Turkish_Northwest_Scaled
0.01366012 Turkish_Aydin
0.01429300 Turkish_Southwest
0.01586123 Turkish_Balikesir
0.02141326 Turkish_North
0.02318909 Turkish_South
0.03119369 Turkish_Central
0.03628459 Turkish_Istanbul
0.03994909 Turkish_Adana
0.04509799 Turkish_Kayseri
0.05295727 Kabardin
0.05312276 Kumyk
0.05421040 Azeri
0.05744611 Abazin
0.05844781 Azeri_Dagestan
0.05871224 Karachay
0.05928395 Cherkes
0.05933765 Azeri_Turkey
0.05970684 Circassian
0.06407075 Balkar
0.06500741 Turkish_Balkans
0.07378559 Talysh_Azerbaijan
0.07407474 North_Ossetian
0.07458294 Adygei
0.07459124 Ingushian
0.07477122 Iranian_Fars



Distance to: Turkish_Northwest_Unscaled
0.00776032 Turkish_Aydin
0.00827017 Turkish_Southwest
0.00891742 Turkish_Istanbul
0.00924031 Turkish_South
0.01014725 Turkish_Central
0.01028508 Turkish_Balikesir
0.01149685 Turkish_North
0.01166097 Turkish_Adana
0.01241129 Turkish_Kayseri
0.01893877 Turkish_Balkans
0.01894235 Azeri
0.02007283 Greek_Central_Anatolia
0.02023798 Greek_Kos
0.02042966 Greek_Crete
0.02075489 Turkmen
0.02092941 Armenian
0.02105170 Ashkenazi_Belarussia
0.02120446 Greek_Izmir
0.02123976 Ashkenazi_Lithuania
0.02182437 Ashkenazi_Russia
0.02269943 Georgian_Jew
0.02275412 Talysh_Azerbaijan
0.02282242 Ezid
0.02284440 Azeri_Dagestan
0.02291654 Italian_Basilicata


Moreover, despite I recommend to use unscaled in single population sharing of Turks, scaled is still a better choice most of time while looking for admixture breakdown.




I'm around somewhere around 15% East Eurasian (mostly ASI with a small amount of Mongoloid) but scaled is much better for me.





For scaled, all the Pashtun samples are ranked close to one another at the top while for unscaled, there's less consistency. The scaled also matches better with the results on GEDmatch. I've yet to see samples where unscaled was better than scaled. I don't think there's any overestimation in the distances for scaled, them having extreme differences makes sense.

Yeah, that's why I said Turks. As I have no experience on South Asians.


I've yet to see samples where unscaled was better than scaled

You can check the example I posted above.

Kaspias
07-22-2020, 11:20 AM
Let's evaluate again what happened in this thread:

- A person who doesn't even have idea what either Rumelia or Deliorman samples is claimed something about them, he insisted all the way despite I said this is not true. In the end I have been accused of being agenda.
- I have claimed that unscaled coordinates work better for Turks in single population sharing, and it really does, but I received many irrelevant comments who are from everyone except for Turks which claims scaled coordinates are better for them while I didn't claim it's opposite. In fact, while stated scaled coordinates indeed work better most of the time.

Thank you all for ruining this thread. I'm closing comments, and probably think more than twice next time before contributing both this forum and community.