PDA

View Full Version : USC admixCalc, GPS Tool and reAdmix^2



Telfermagne
01-09-2015, 03:17 AM
Here I am taking a look at the University of Southern California’s Geographic Population Structure Prediction tool. I am curious what y'all think of this tool: http://chcb.saban-chla.usc.edu/gps/index.php.

From what I can tell it's consistent enough with 23andMe's Ancestry Composition, Dr. McDonald's BGA and SPA UCLA.

First I ran my 23andMe raw data file through admixCalc, which can be found at http://chcb.saban-chla.usc.edu/admixCalc/index.php/home. Second I put my admixCalc outcomes into GPSP’s Step 1.

During Step 2 one must select an analysis option based upon how diverse one’s ancestry is. If one has a very diverse background, then one ought to choose reAdmix². If one does not have a very diverse background, then one ought to choose GPS Tool. I will work with both just to see what happens.

For GPS Tool simply select it and click proceed to next step. My outcome said that I am nearest to Belgians. An outcome from reAdmix² took a bit longer to get; first, I was asked to choose population descriptors that pertain to what I suspect about my ancestral background. Even though I have done genealogical research before, both traditional and genetic, I have adopted a mildly skeptic mindset. So I selected zero populations. After choosing from the population descriptors there is a confidence selection pertaining to how confident one is in the accuracy of the given information, e.g. Very Confident, Not Confident, or Unsure/Not So Confident. I selected Unsure/Not So Confident.

It would seem that one ought to not select zero populations as after about half an hour of waiting I got no results. For my second try at reAdmix² I chose populations that were consistent with the major components reported by previous admixture outcomes. Given what I have witnessed on various online discussions pertaining to admixture outcomes I think that I ought to explain what I mean by “major components”. In my mind a major component would be something like 27% Martian-Face-Eater, while 0.54% Plutonian-Virgin-Catcher would not be a major component. The major components/populations that I chose were Anglo-Saxon, British, German, Highland Scottish, Irish, and Welsh. As for my confidence selection, I again chose Unsure/Not So Confident.

After around two and a half minutes my results were ready. My scores were as follows:

27.7% Welsh
21.5% Luxembourgish
18.4% British
10.7% Bavarian
7.5% Highland Scottish
7.1% Iberian
4.3% Ashkenazi
2.2% Kuwait
0.6% Abkhazian

Lastly I substituted each admixture component with a set of coordinates, and I recycled the percentages as weights pulling between the coordinates. The goal was to find a rough weighted midpoint so to see where my reAdmix² results stand with regard to my GPS Tool outcome.

As was my case in the "Mapping Your Location (Gedmatch calculators)" thread, I ran into some trouble with regard to how to consider the Ashkenazi. Even though places like Poland have a history of high Ashkenazi populations, I don’t think that using Poland’s coordinates is appropriate as there is a distinct “Polish people” that can be alternatively associated with Poland’s coordinates. The Ashkenazi seem to be ultimately connected to the Fertile Crescent, so I might as well put in Israel even though they may bear some traces of the folk they settled around in Europe.

The outcome was remarkably close to what I got with GPS Tool, they were well within 100 miles of each other; to be more precise Rilland, Netherlands is roughly 74 miles from Belgium. The outcome is also near exact to those of the same kind that I got from breaking down my 23andMe results, Dr. McDonald BGA results, and UCLA SPA results. Therefore, I am obliged to think highly of USC's tool.

I plugged the following into geomidpoint, http://www.geomidpoint.com/:
*selected multiple input & other weight.

United Kingdom of Great Britain = 0.536
Luxembourg = 0.215
Bavaria = 0.107
Spain = 0.071
Israel = 0.043
Kuwait = 0.022
Abkhazia = 0.006

Outcome: Rilland, Zeeland, Netherlands. Roughly 74 miles from Belgium.

Telfermagne
01-10-2015, 07:47 PM
Did it a second time as the first time I forgot to add Austrian & Bavarian to the population selection list. The second set of major components/populations that I chose were Anglo-Saxon, Austrian, Bavarian, British, German, Highland Scottish, Irish, Scottish, Swiss, and Welsh. This time around I dropped the "skeptic" attitude and chose a “Very Confident” confidence level as the selected groups are consistent with the results of previous investigations.

It would seem that the results shift based upon the confidence level one chooses. If one seem sure or near sure that one is from a certain set of populations, then the specificity of the outcome is biased to favor that sureness. What seems to matter is where the various outcomes average out at, and this outcome averages out near to previous ones and as such I see no reason to fault it.

This time around I got 53.3% British_0 and 46.7% Bavarian_0

vettor
01-10-2015, 08:17 PM
Did it a second time as the first time I forgot to add Austrian & Bavarian to the population selection list. The second set of major components/populations that I chose were Anglo-Saxon, Austrian, Bavarian, British, German, Highland Scottish, Irish, Scottish, Swiss, and Welsh. This time around I dropped the "skeptic" attitude and chose a “Very Confident” confidence level as the selected groups are consistent with the results of previous investigations.

It would seem that the results shift based upon the confidence level one chooses. If one seem sure or near sure that one is from a certain set of populations, then the specificity of the outcome is biased to favor that sureness. What seems to matter is where the various outcomes average out at, and this outcome averages out near to previous ones and as such I see no reason to fault it.

This time around I got 53.3% British_0 and 46.7% Bavarian_0

I seen this program before , about 2 years ago in beta form
Eurogenes created for gedmatch the program K9b for it................

The old beta one mark1 had me as Central france,.......mark2 beta placed me in hungaria
This one places me in gotenburg Sweden

using the manual part to select your origins and select very confident ....I was given an equal split of Finnish, Basque and Yukaghir

........................

My son get placed 100Km south of gothenbeurg

and manual input test, has 27% north-Italian and the rest bermudian ( I have no clue what this represents )

.....................

for wife..............she was placed in aalborg denmark
and the manual split is 14% north-italian, 19% french, 28% slovak and the rest yukaghir


I see a northern trend for this program and I wonder if this "bermudian" means unsure


...........EDit- I use the geomidpoint and got the following

my wife = east gotland

me 23andme ( v3 )data = Krzykaly in Poland ( near gdansk )
me ftdna data = Rocroi in NE France ( near belgium )

All I can say is that the slighest numbers makes a big difference to this mapping..........which is why 23andme and ftdna make it different

Telfermagne
01-18-2015, 11:12 PM
Did a third run. This time I included populations that coincide with not only previous genealogical knowledge but also Relative Finder & Family Finder matches which revealed previously unknown links to Ashkenazi Jews, Poles, Ukrainians, Russians (within 5 to 7 generations) and more distantly Croats. If I had to guess the matches from Poland and Russia have to do with some maternal ancestry connected to Danzig and Königsberg (coincidentally these matches tend to be on the X-Chromosome).

Populations: Anglo-Saxon, Ashkenazi, Austrian, Bavarian, British, Croatian, German, Highland Scottish, Irish, Polish, Russian, Scottish, Swiss, Ukrainian, Welsh.
Confidence level: Very confident (the populations selected conform to what has been revealed by Relative Finder & Family Finder matches and genealogies).

Outcome:

51.1% British
45.8% Bavarian
3.1% Ashkenazi

Dr_McNinja
01-19-2015, 11:42 AM
The reAdmix^2 tool didn't give me coherent results.

The normal reAdmix tool gave me


Ancestral Route 1 India_Uttar_Pradesh_Brahmins_0 26.06 83.18
Ancestral Route 2 Panjabi_0 30.79 76.78

Pic: http://i.imgur.com/DUYFaq3.png

And the GPS gave me different results from last time I used this tool. Last time it put me and anyone remotely like me somewhere in Rajasthan near the border. This time it said Punjabi Sikh and put me here:


Based on the data that you have entered our tool computed your closest population to the Panjabi Sikh region.
Calculated Latitude of Origin: 28.8684
Calculated Longitude of Origin: 78.21

Pic: http://i.imgur.com/E2cH5jR.png

reAdmix^2 when selecting Panjabi Sikh and 'Unsure' gave me 80.5% Panjabi Sikh with nothing else.

That was using my Geno 2.0 results which were:

Southwest Asian: 50%
Mediterranean: 16%
Southeast Asian: 19%
Northeast Asian: 2%
Native American: 2%
North European: 11%

But when I used their admixture tool, it gave me:

56.15% SW-Asian
16.28% Mediterranean
21.16% SE-Asian
0.94% NE-Asian
0.64% Native American
4.83% North European

Which is basically similar but 6% swapped form North European to SW-Asian. That also is similar somewhat to the Geno 2.0 results for South Indians (see here: https://docs.google.com/spreadsheet/ccc?key=0AuXBmvmgdkfVdFMtRHVlZDBuQ3lMcjhxMDE4V3JoY lE#gid=11 )

When that is put into the tool, it gives me that old Rajasthan area again. The reAdmix gives me Rajasthan + South India.

It looks like they fed in our reported Geno 2.0 results into their references?

Sein's Geno 2.0 results give something like India and South America for reAdmix which I thought was very surprising and interesting.

FWIW, HRP0370's admixture results as calculated by their tool:

47.82% SW-Asian
33.77% Mediterranean
8.81% SE-Asian
1.81% NE-Asian
0.88% Native American
6.90% North European

They put that on the Afghan-Pak border, on the Afghan side, and said it was Baloch. In reAdmix, they came out as 100% Abhkazian over and over (except the map showed 30% or something, so it didn't make sense... I played it with until it gave 20-30% Abhkazian and 70-80% Pashtun).

Dr_McNinja
01-19-2015, 12:07 PM
So you can select individual populations in reAdmix^2, just to get that proportion with nothing else. So I got 7.3% Orcadian, then 16.5% Finnish, then I tried selecting Basque and it flipped out on me and gave me the attached results. 18.5% Mordovian, 17.8% Luxembourgish, 63.6% Piramalai Kallar which was from 53.2 Piramalai Kallar, 15.5 Mordovian, 14.9 Luxembourg, 16.4 Unknown.

Then I tried Abhkazian and got 69.5% Mogaveera, 15.4% Austrian, 14% Yukagir (Siberia/Northeast Asia?), 1.1% Apatani. The numbers not in parenthesis gave 57 Mogaveera, 12.6 Austrian, 11.5 Yukagir, 0.9 Apatani, 18 Unknown.

I got 11.9% for Balochi, then I tried selecting India Pulliyar and it gave me 15.9% Austrian, 10.9% Georgian, 4.1% Yukagir, 69.1 Unknown. I tried Vietnam and it gave me 91.8% Uttar Pradesh Brahmin, 3.8 pygmy.

Salkin
01-19-2015, 12:23 PM
The GPS tool places me in western Denmark. That's not completely off base, I guess, though most tools consider me more of a Norwegian.

The reAdmix (not 2) puts me a little further east, on the eastern shore of Jylland/Jutland vs. GPS on the western shore.

For reAdmix2, I selected Swedish and Belgian (the Belgian influence is very minor, but the only non-Swedish admixture I'm aware of having), Very Confident, and it called me 100% Belgian. I re-ran it with only Swedish selected and came up 100% Swedish. At that point I went "huh, is it just going to take one of the components I indicate and confirm it regardless?" and tried a run with Vietnamese + Not Confident. That gave me 95.4% Icelandic + 3.7% Iberian + 0.9% Askham_Northern_Cape_RSA_0, whatever that is.

Dr_McNinja
01-19-2015, 12:26 PM
Messing around with Sein's I managed to get 22.3% Kapu (South India), 77.1% Kalash, 0.6% Adygei. It also kept giving 13-15% Abkhazian, but wouldn't give a Finnish score. After some tinkering, I finally got an Eastern European score, 8.6 Lithuanian, 2.9 Yukagir, 72.3 Brahui, 16.2 Unknown.

So some of these numbers do come out reasonably close to the other Admixture calculators.

The readmix for HRP0393 (Haryana Jatt) gives Punjab and Rajput_Rajasthan which is interesting. Their admixture results (not Geno 2.0 but that USC site):

53.43% SW-Asian
17.96% Mediterranean
17.07% SE-Asian
1.01% Native American
10.42% North European

Got 35 Brahui, 25.7 Kalash, 6.1 Lithuanian, 33.2 Unknown for HRP0393's numbers there. That 35% Brahui is close to their Harappa Baloch.

Using Paulgill's Geno 2.0:

Southwest Asian 50%
Southeast Asian 15%
Mediterranean 19%
North European 13%
Northeast Asian 2%

I got 11.5% Finnish (his Harappa NE-Euro is 11.33%), fooling with the confidence settings got it up to 15.8% (mine ranged from 15.1 to 16.5 which is odd since my Geno 2.0 numbers have lower European all around... I'm guessing the Native American must've made a difference).

Telfermagne
01-21-2015, 02:49 AM
The GPS tool places me in western Denmark. That's not completely off base, I guess, though most tools consider me more of a Norwegian.

The reAdmix (not 2) puts me a little further east, on the eastern shore of Jylland/Jutland vs. GPS on the western shore.

For reAdmix2, I selected Swedish and Belgian (the Belgian influence is very minor, but the only non-Swedish admixture I'm aware of having), Very Confident, and it called me 100% Belgian. I re-ran it with only Swedish selected and came up 100% Swedish. At that point I went "huh, is it just going to take one of the components I indicate and confirm it regardless?" and tried a run with Vietnamese + Not Confident. That gave me 95.4% Icelandic + 3.7% Iberian + 0.9% Askham_Northern_Cape_RSA_0, whatever that is.

The more I screw around with this tool, the more that it seems such is the case. The various combinations I get still "average" out, but they seem unreliable in the sense of the bolded & quoted.

ZephyrousMandaru
01-21-2015, 05:52 AM
What raw data files are you guys using? I haven't been able to do it.

vettor
01-21-2015, 05:56 AM
What raw data files are you guys using? I haven't been able to do it.

Eurogenes K9b

and total when data is entered should not exceed 1.0

so as an example, a 56.33...should be entered as 0.5633

vettor
01-21-2015, 05:58 AM
The more I screw around with this tool, the more that it seems such is the case. The various combinations I get still "average" out, but they seem unreliable in the sense of the bolded & quoted.

I think the default is best and seems to be maintained in some degree with the tests below

Salkin
01-21-2015, 08:30 AM
What raw data files are you guys using? I haven't been able to do it.

I used my 23andMe .txt file. I didn't see anything on the site about that format being accepted, but someone in the thread mentioned it.

DMXX
01-21-2015, 09:58 AM
Overall, a very disappointing experience.

Using the GPS tool, I decided to "correct" my Geno 2.0 results (http://www.anthrogenica.com/showthread.php?1066-DMXX-s-Geno-2-0-Results) by proportionally recalibrating the scores, as the total adds up to 98%. All I did was divide each score by 0.98. This bypasses the issue of arbitrarily making up the difference by adding a percentage point here and there. Example of what my scores looked like in the Spoiler below.


NE Asian 0.05 0.051020408
Med 0.47 0.479591837
SW Asian 0.32 0.326530612
SE Asian 0.02 0.020408163
N Euro 0.12 0.12244898


The tool placed me in Northeast Turkey. This isn't too far from my family's ancestral region (NW-N Iran), but for a program being touted as an accurate designator of biogeographic origins, it's worse than the more rudimentary methods we're all familiar with (e.g. McD's BGA).

The Re-Admix Tool placed me even further north (Maikop territory). Which is pretty bizarre.

3530

The Re-Admix2 tool produced similar results. I noticed they've erroneously listed Tajiks, Uzbeks and Balochis as European, yet Black Sea Turks are Asian. Despite attempting all combinations of "Iranian", "Persian" and "Azerbaijani", the results are similar each time. I'm apparently predominantly Abkhazian:

3531

The Abkhazians cannot be considered a surrogate for generic West Asian ancestry given most West Asian ethnicities are represented here (there's even a half-dozen variants of Lebanese!). If reference population count isn't an issue here, then something with their methodology is fundamentally flawed.

[Edit]: Attempted this again using my Eurogenes K9b results from GEDmatch. Results are as follows:
GPS - Iranian-Afghan border
Re-Admix - Described as "Tajik-Lezgin route"
Re-Admix2 - Using "Persian" and "Azerbaijani", the program's refuting my Azeri ancestry and saying I'm 55.8% Persian and 44.2% Yukaghir (http://en.wikipedia.org/wiki/Yukaghir_languages). This is one of the strangest results I've ever seen.

AJL
01-21-2015, 04:12 PM
I think I've broken the utility.

Using K9b, GPS gives me a little northeast of Gothenburg, Sweden. It's a region I probably have links to but there's no way it is an acceptable average of my total ethnic makeup.

reAdmix with two-parent scenario (the other one doesn't seem selectable) gives me Finland and Bermuda.

v2 gives me 100% Yukaghir on single mode so I suspect there is something the matter with K9b, with Gedmatch's version of it, or most likely the Yukaghir sample set as well as the whole calculator. Whatever result from my known ancestry I select, I get strange combinations, e.g.:

German 34.8
Scandinavian 33.9
Finnish 22.5
French 6.6
Gumuz 2.3

or

Anglo-Saxon 32.25
Biaka Pygmy 40.3
Bermudan 26.1
Yukaghir 1.4

or

Highland Scottish 46.7
Tatar 34.0
Basque (French) 13.4
Auak 5.8

or

Carpatho-Rusyn 32.5
Belarussian 22.3
Yukaghir 21.7
Sephardic Jews 10.4
Iberian 9.0
Vanuatu 4.2

I suspect there's some problems with their mapping since my mother is in the North Sea, and on split mode gets Yukaghir and Lithuanian. My sister, whose results just came in, is also "Finnish/Bermudan," and in multiple mode gets similarly whacky results.

DMXX
01-21-2015, 04:19 PM
v2 gives me 100% Yukaghir on single mode

Well, there went my hopes I'm a special Turkic snowflake displaced from the Siberian steppe I guess.

AJL
01-21-2015, 04:26 PM
Well, there went my hopes I'm a special Turkic snowflake displaced from the Siberian steppe I guess.

Unless there was mass Siberian migration to northwest England, the Hebrides, Provence, and the Rhineland as well, probably so!

vettor
01-21-2015, 04:51 PM
I see the program as worse than its previous 2 beta programs

I can only imagine that they use Yukagiur ( siberian ) for an unknown "somewhere in the ancient eastern area " and also use Bermudian for an unknown "somewhere in the ancient western area ".
maybe the program as most of these types of programs only goes back 2000 years........

ignore these 2 and it might be some credence to one's origins

So, my family's results would be

Me = Finnish and Basque ...........must mean very very old in europe .....

my wife = 14% north-italian, 19% french, 28% slovak

my son = 27% north-Italian

.......

MfA
01-21-2015, 05:18 PM
Running my raw data through USC's admix tool:


59.63% Mediterranean
37.16% South West Asian
2.59% North East Asian
0.61% North American
0.00% North European
0.00% Oceanian
0.00% South African
0.00% Southeast Asian
0.00% Sub Saharan African

http://abload.de/img/desktop_2015_01_21_19meijz.png

Ezidi Kurd


57.12% Mediterranean
40.16% South West Asian
2.15% North East Asian
0.56% North American
0.00% North European
0.00% Oceanian
0.00% South African
0.00% Southeast Asian
0.00% Sub Saharan African
http://abload.de/img/desktop_2015_01_21_19d1czk.png

ZephyrousMandaru
01-22-2015, 04:20 AM
Northeast Asian 0.00%
Mediterranean 66.45%
South African 0.00%
Southwest Asian 33.54%
North American 0.00%
Oceanian 0.00%
Southeast Asian 0.00%
North European 0.00%
Sub-Saharan African 0.00%

My GPS

http://oi57.tinypic.com/30mpvnr.jpg

My reAdmix results

http://oi57.tinypic.com/30w1fh3.jpg

My reAdmix 2 results

http://oi59.tinypic.com/dwf0qt.jpg

85% Lebanese_Beirut_Christian, 13.1% Yemenese_1 and 1.9% BiakaPygmy_1

Humanist
01-22-2015, 04:27 AM
My GPS

Just a tad bit off. :P

ZephyrousMandaru
01-22-2015, 04:56 AM
Just a tad bit off. :P

Yeah I know, the reAdmix 2 tool yields different results, depending on which populations you select as your "ancestral" population. I chose Iraqi Jews, this time I received 86.6% Lebanese_Beirut_Christian, 13.3% Yemenese_1 and 0.1% Bangledeshi. If I had more of the Southwest Asian component, I'd probably cluster eastwards towards Iraq.

vettor
01-22-2015, 05:33 AM
I ran Ust numbers and got

closest population to the Andamanese region.
near town of Nellore in East India


Hinxton 4 came out as in Doggerland


malta boy ( ma-1) came out as
closest population to the Suomalainen region.
near tallinn, estonia


and lastly ..a female, hungarian CO2
closest population to the Iberian region.
near Nancy, France

Roaring
01-22-2015, 08:48 AM
GPS places me surprisingly well, considering that im either 3\4 Russian 1\4 German, or 7\8, 1\8.

http://s018.radikal.ru/i521/1501/5d/5718427346fc.png (http://www.radikal.ru)

Two pops don't catch any of my ancestors populations, but if we will accept that North-Russians are close to Finns, and Norwegians are similar to Germans, well it can make some sense.

http://s017.radikal.ru/i407/1501/b6/66b0cf95f9ce.png (http://www.radikal.ru)

DMXX
01-22-2015, 09:13 AM
I ran Ust numbers and got

closest population to the Andamanese region.
near town of Nellore in East India


Hinxton 4 came out as in Doggerland


malta boy ( ma-1) came out as
closest population to the Suomalainen region.
near tallinn, estonia


and lastly ..a female, hungarian CO2
closest population to the Iberian region.
near Nancy, France

What about IR1, vettor? Do you have his genome?

vettor
01-22-2015, 09:29 AM
What about IR1, vettor? Do you have his genome?

Here you are , all that gedmatch accepted

http://www.y-str.org/p/ancient-dna.html


some good genetic tools as well in the link

jesus
01-22-2015, 10:14 AM
Overall, a very disappointing experience.

Using the GPS tool, I decided to "correct" my Geno 2.0 results (http://www.anthrogenica.com/showthread.php?1066-DMXX-s-Geno-2-0-Results) by proportionally recalibrating the scores, as the total adds up to 98%. All I did was divide each score by 0.98. This bypasses the issue of arbitrarily making up the difference by adding a percentage point here and there. Example of what my scores looked like in the Spoiler below.


NE Asian 0.05 0.051020408
Med 0.47 0.479591837
SW Asian 0.32 0.326530612
SE Asian 0.02 0.020408163
N Euro 0.12 0.12244898


The tool placed me in Northeast Turkey. This isn't too far from my family's ancestral region (NW-N Iran), but for a program being touted as an accurate designator of biogeographic origins, it's worse than the more rudimentary methods we're all familiar with (e.g. McD's BGA).

The Re-Admix Tool placed me even further north (Maikop territory). Which is pretty bizarre.

3530

The Re-Admix2 tool produced similar results. I noticed they've erroneously listed Tajiks, Uzbeks and Balochis as European, yet Black Sea Turks are Asian. Despite attempting all combinations of "Iranian", "Persian" and "Azerbaijani", the results are similar each time. I'm apparently predominantly Abkhazian:

3531

The Abkhazians cannot be considered a surrogate for generic West Asian ancestry given most West Asian ethnicities are represented here (there's even a half-dozen variants of Lebanese!). If reference population count isn't an issue here, then something with their methodology is fundamentally flawed.

[Edit]: Attempted this again using my Eurogenes K9b results from GEDmatch. Results are as follows:
GPS - Iranian-Afghan border
Re-Admix - Described as "Tajik-Lezgin route"
Re-Admix2 - Using "Persian" and "Azerbaijani", the program's refuting my Azeri ancestry and saying I'm 55.8% Persian and 44.2% Yukaghir (http://en.wikipedia.org/wiki/Yukaghir_languages). This is one of the strangest results I've ever seen.

Lol you're not alone.
Eurogenes K9b results:
http://i61.tinypic.com/4zzr5w.png

AJL
01-22-2015, 04:03 PM
I've been thinking some more about the GPS and am wondering why this tool is so terribly off.

So then I decided to see where each component was centred by doing each at 100%.

SW Asian: near Hyderabad, India
NE Asian: Mongolian/Chinese border
South African: northern South Africa
Native American: Peru/Bolivian border
Oceanian: Papua New Guinea
SE Asian: NE India
North Euro: North Sea, a little east of Edinburgh
Subsaharan African: western Ethiopia

Since neutral (non-editorializing) creation of clusters should pick one or more West Asian/Caucasus/Eastern Mediterranean clusters out of the first nine, rather than two different South Asian ones (http://dienekes.blogspot.ca/2010/12/human-genetic-variation-first.html) a reasonable conclusion is that either they have wrongly centred SW Asia, or their choice of clusters is very poor -- or both.

MitchellSince1893
01-22-2015, 04:10 PM
I have to say this result matches my genealogical research the best of all admixtures I've used (every one on gedmatch) , McDonald, and SPA.

Genealogically speaking (not what I actually genetically inherited from my ancestors) I'm 82% British Isles. 15% German, French, and Dutch, 2% Scandinavia, 1% India.

GPS places me at 52°08'27.2"N 1°05'03.7"W near Blacksley, England

If I simply use the Geographic Midpoint Calculator http://www.geomidpoint.com/ entering the country level genealogical percentages (very rough locations) I come out at:

53.440493 N, 0.224051E near Saltfleet, England.

These 2 pts are 105 miles apart.

I will create a map later showing cntr points for various admixture calcultors.

AJL
01-22-2015, 04:26 PM
I have to say this result matches my genealogical research the best of all admixtures I've used

You know what they say about a stopped clock, right?

I tested a cousin who's mainly Isles with a little German, Swiss and Italian. She should end up in France, but instead winds up near Malmö, Sweden.

My mother places, with every reasonable tool, between the English Channel and Aachen area. With this she is in the Baltic Sea.

All the results I have seen are too northeast-shifted.

AJL
01-22-2015, 05:04 PM
I just realized I made a major omission, which was not seeing where the Mediterranean component was centred. It's in Cairo.

CelticGerman
01-22-2015, 06:48 PM
Tried the most simple version for 4 persons using EUROGENES K9b percentages. Plot in North Sea for all four. This could be correct for two of them compared to results from McDonald, David W. or Interpretome. But for the two others it's impossible. A person from France (Burgundy and Rhône-Alpes) with an average close to the Danish coast??? This is definitely wrong. A central German profile close to Scotland, same as a person with clearly Northern German/Scandinavian profile. Which means I stop loosing time with this tool.

MitchellSince1893
01-23-2015, 12:08 AM
You know what they say about a stopped clock, right?

I tested a cousin who's mainly Isles with a little German, Swiss and Italian. She should end up in France, but instead winds up near Malmö, Sweden.

My mother places, with every reasonable tool, between the English Channel and Aachen area. With this she is in the Baltic Sea.

All the results I have seen are too northeast-shifted.

Here is a sampling of my admixture results locations. As you can see GPS is the most Western for me.

3556

ZephyrousMandaru
01-23-2015, 12:38 AM
I just realized I made a major omission, which was not seeing where the Mediterranean component was centred. It's in Cairo.

That explains why all of the populations I chose as "ancestral" to me, where almost always centered in the Southern Levant region, as well as in Arabia.

My reAdmix 2 results after selecting every Middle Eastern, North African and West Asian populations.

http://oi61.tinypic.com/1zefuqv.jpg

AJL
01-23-2015, 01:08 AM
Here is a sampling of my admixture results locations. As you can see GPS is the most Western for me.

3556

Maybe so, but we already have 5 counterexamples from a very small pool. That speaks to exceedingly high margins of error and a distinct bias northeastwards, which means very low overall accuracy. If you used the actual TGP data, it is of course possible there is a distinct difference between the Gedmatch K9b version of the tool and the Genographic version, since all the bad reports so far are coming from K9b.


That explains why all of the populations I chose as "ancestral" to me, where almost always centered in the Southern Levant region, as well as in Arabia.

My reAdmix 2 results after selecting every Middle Eastern, North African and West Asian population.

http://oi61.tinypic.com/1zefuqv.jpg

It's a major problem for any Near Easterner if a tool given nine components centres two of these components in South Asia, one in Ethiopia, and one in Egypt. Essentially this gives you an axis of nearly half (4/9) of available components running in a narrow band from northeast Africa through the Arabian Peninsula and across the Arabian Sea to India.

Worse, the pull eastward to the two South Asian components is in the same direction, east-southeast, as the pull toward either Native American or Oceanian, relative to the Near East.

That means this entire utility supposes there is essentially one giant population axis running from the UK to Peru, and the only way to move from this axis is south (South Africa) or northeast (NE Asia). I believe this accounts for the northeastward bias seen in multiple kits: if you want to deviate from this arbitrarily constructed axis and it is not an obviously African deviation, you have to do it with a skew northeastward. This would tend to make the tool only work for people whose ancestral populations are right on this axis. This seems like an insurmountable design flaw.

vettor
01-23-2015, 01:23 AM
Here is a sampling of my admixture results locations. As you can see GPS is the most Western for me.

3556

great job on your map

I have done similar , I am not as gifted as you, see below,

had to include legend on left side

Not: these are only my Ftdna data and not my 23andme ( v3 ) data

http://i103.photobucket.com/albums/m153/vicpret/audna233_zpsabc3cd50.jpg (http://s103.photobucket.com/user/vicpret/media/audna233_zpsabc3cd50.jpg.html)


The GPs is this program for my Ftdna .................my 23andme placed me in Poland

Shaikorth
01-23-2015, 01:25 AM
There is a difference between Geno 2 results and K9b results in the GPS tool, K9b being much less accurate.

I don't think this can put an Irishman in Ireland no matter what results are used (just like in PCA's, Isles people don't deviate enough west from the Spain-Sweden line), but it can put a Swede in Sweden.

vettor
01-23-2015, 01:53 AM
There is a difference between Geno 2 results and K9b results in the GPS tool, K9b being much less accurate.

I don't think this can put an Irishman in Ireland no matter what results are used (just like in PCA's, Isles people don't deviate enough west from the Spain-Sweden line), but it can put a Swede in Sweden.

maybe you are correct
my natgeno (not using K9b) ...results are

tool computed your closest population to the French (Basque) region.
Calculated Latitude of Origin: 41.9079
Calculated Longitude of Origin: 3.14334

Near Girona in northern Spain

MitchellSince1893
01-23-2015, 02:05 AM
Maybe so, but we already have 5 counterexamples from a very small pool. That speaks to exceedingly high margins of error and a distinct bias northeastwards, which means very low overall accuracy. If you used the actual TGP data, it is of course possible there is a distinct difference between the Gedmatch K9b version of the tool and the Genographic version, since all the bad reports so far are coming from K9b.

I loaded my 23andme data into admixcalc site and took those component values and put them in the GPSP Admixture Coefficients Entry Form (step 1), and for step 2 I used the GPS Tool.

The clock analogy may be accurate for me...maybe it was just been dumb luck that it came out in England. My father came out near Luxembourg and my mother in the North Sea near the Frisian coast above the Netherlands.

My father should have been Northwest of me...instead he was southeast. My mom's results seemed reasonable.

jesus
01-23-2015, 02:14 AM
I uploaded my Raw data to their website and I got way more accurate results in comparison with Eurogenes K9b. I plot in Semnan Province, Which confirms my Parthian Horse archer ancestry.

Source: http://en.wikipedia.org/wiki/Semnani_language

Map of Parthia http://3.bp.blogspot.com/-fzvs9NyKzXo/T_C7bcysS-I/AAAAAAAAAcQ/vZYUmQOMoDA/s1600/Encyclopaedia+Britannica+Parthian+Empire.png


JK :P

But the flag of the Parthian empire should be added imo


Based on the data that you have entered our tool computed your closest population to the Iranian region.
Calculated Latitude of Origin: 34.3772
Calculated Longitude of Origin: 53.4491

http://i59.tinypic.com/e8l6j5.jpg

Sein
01-23-2015, 04:02 AM
In my case, the GPS tool was extremely accurate. I entered my actual Geno 2.0 results, in this form:

Southwest Asian=0.4625
Mediterranean=0.2525
Northern European=0.1125
Southeast Asian=0.1125
Northeast Asian=0.04
Native American=0.02

I just added .25 to my four largest components.

Last time, it put me in northeastern Iran. Not necessarily inaccurate in terms of broad global genetic affinities, but completely wrong in terms of actual genealogical roots. Now, it has me closest to the Kalash, and I plot at 34°50'27.2"N, 71°05'29.4"E.

3558

That's in the neighborhood of Asadabad, in the Kunar province of Afghanistan (and very close to the Durand Line). Out of all Pashtuns, I am most genetically similar to Pashtuns from the regions of Kunar, Laghman, and Nangarhar in Afghanistan, and to Pashtuns from Bajaur Agency and Mohmand Agency in Pakistan. Also, most of my actual genealogical ancestry is from this general region. So, this is a rare example (at least for me) of one of these algorithms having me plot in the correct spot, both genetically and genealogically.

The reAdmix tool has me as being intermediate between the Kalash and the HGDP Pashtuns:

3559

Not necessarily bad, considering that my parents aren't from different backgrounds, so it gives me two virtually identical populations. Also, although this might not be related, my mother has an excess affinity to the Kalash population, in comparison to the HGDP Pashtun samples. This is very evident on PCA plots. I think this might be evidence that Pashtun tribes this far up north (my mother is mostly Nangarhari Mohmand, with some Paktiawal Mangal and Khyberi Afridi) probably have some cryptic admixture from Dardic peoples like the Kalash and Pashayi, as well as some admixture from Nuristani groups. So far, very reasonable.

The reAdmix2 tool construes me as being 88% HGDP Pashtun, 6% Lezgin, and 6% Abkhazian. This is after I choose Pathan and Lezgian for my populations (and "Unsure/Not so confident"). I did this mainly because most GedMatch Oracles have me as being 85%-90% HGDP Pashtun, and 15%-10% Lezgian. So, this gave me a result that I'm quite used to. All in all, good stuff.

MitchellSince1893
01-23-2015, 04:31 AM
Sorry for the tangent; but I found this interesting as I used http://www.geomidpoint.com/ to plot out my various admixture results.

My Dodecad K7b, Eurogenes K15, MDLPK23b, Eurogenes EUTest, Eurogenes K13, and Eurogenes Hunter-Gatherer center points polygon area looks a lot like the homeland, route, and destination of the Angles tribe (click on tiny blue map right below).

3560
https://i0.wp.com/upload.wikimedia.org/wikipedia/commons/thumb/5/53/Britain.Anglo.Saxon.homelands.settlements.400.500. jpg/800px-Britain.Anglo.Saxon.homelands.settlements.400.500. jpg

Of course it could also fit Danelaw immigration
http://www.tartanday-wa.org/Images/Vikings.png

Maybe Eurogenes K13 and Hunter-Gatherer are showing more ancient origins, while the others are more recent in time.

Sein
01-23-2015, 04:35 AM
I ran Paul Gill's results (Sikh Punjabi Jatt):

3561

Right in the middle of Indian Punjab, which is pretty genealogically accurate.

bored
01-23-2015, 05:04 AM
This tool placed me in Maharashtra lol, which is way off the mark.

jesus
01-23-2015, 05:15 AM
My sister's result, she's usually 2-3% different than me on Gedmatch calculators. She's in Northeast of Markazi Province, in the Khalaj territory.
http://en.wikipedia.org/wiki/Khalaj_language_(Turkic)

http://i59.tinypic.com/k31x05.jpg

bored
01-23-2015, 05:15 AM
Eurogenes K9b is one of the worst calculators out there IMO. I gives me <1 % NE Asian and puts my East Eurasian into Southeast Asian instead. Some ASI also goes into Southeast Asian. It's not consistent though. Sapporo, for example, gets higher NE Asian than me even though I have higher actual East Eurasian. But that ends up lowering his Southeast Asian.
If most NW South Asians get low Southeast Asian then the GPS tool will plot me way further south because for me, Eurogenes confounds my East Eurasian with "Southeast Asian"( which is just ASI for most other NW SAsians). Overall, terrible calculator.

Sein
01-23-2015, 05:20 AM
This tool placed me in Maharashtra lol, which is way off the mark.

That's probably because you used your K9b results, which is just too different from Geno 2.0. When I use those results, I end up in Rajasthan.

jesus
01-23-2015, 05:29 AM
That's probably because you used your K9b results, which is just too different from Geno 2.0. When I use those results, I end up in Rajasthan.

My K9b results placed me close to Zahedan, on the Pakistani-Iranian-Afgan Border. It also placed my sister in Afghanistan lol. http://en.wikipedia.org/wiki/Zahedan

ZephyrousMandaru
01-23-2015, 05:50 AM
The only reAdmix population combinations that come close to accurate. Populations selected were Iranian Jews, Iraqi Jews, Iraqis, Azerbaijani Jews, Georgian Jews and Armenians.

http://oi58.tinypic.com/jr8ieg.jpg

vettor
01-23-2015, 06:19 AM
I uploaded my Raw data to their website and I got way more accurate results in comparison with Eurogenes K9b. I plot in Semnan Province, Which confirms my Parthian Horse archer ancestry.

Source: http://en.wikipedia.org/wiki/Semnani_language

Map of Parthia http://3.bp.blogspot.com/-fzvs9NyKzXo/T_C7bcysS-I/AAAAAAAAAcQ/vZYUmQOMoDA/s1600/Encyclopaedia+Britannica+Parthian+Empire.png


JK :P

But the flag of the Parthian empire should be added imo


Based on the data that you have entered our tool computed your closest population to the Iranian region.
Calculated Latitude of Origin: 34.3772
Calculated Longitude of Origin: 53.4491

http://i59.tinypic.com/e8l6j5.jpg

how did to upload your data, are you registered?

vettor
01-23-2015, 06:27 AM
Eurogenes K9b is one of the worst calculators out there IMO. I gives me <1 % NE Asian and puts my East Eurasian into Southeast Asian instead. Some ASI also goes into Southeast Asian. It's not consistent though. Sapporo, for example, gets higher NE Asian than me even though I have higher actual East Eurasian. But that ends up lowering his Southeast Asian.
If most NW South Asians get low Southeast Asian then the GPS tool will plot me way further south because for me, Eurogenes confounds my East Eurasian with "Southeast Asian"( which is just ASI for most other NW SAsians). Overall, terrible calculator.


I tend to agree with you about Eurogenes K9b

vettor
01-23-2015, 06:31 AM
The only reAdmix population combinations that come close to accurate. Populations selected were Iranian Jews, Iraqi Jews, Iraqis, Azerbaijani Jews, Georgian Jews and Armenians.

http://oi58.tinypic.com/jr8ieg.jpg

sometimes I get numbers after the ethnicity , as an example on yours, you have Iranian jews _0...............I get stuff like Abkazian_3, or French_1

Any idea what this might mean?

jesus
01-23-2015, 06:43 AM
how did to upload your data, are you registered?

first you register here and Upload your raw data here http://chcb.saban-chla.usc.edu/admixCalc/index.php/home then after you're done (It takes few minutes but it can faster sometimes) click on compute. Then it would show you your admixture results.

After that you have to plug the results in here http://chcb.saban-chla.usc.edu/gps/index.php

vettor
01-23-2015, 06:51 AM
The program is very good when I use my natgeno

http://i103.photobucket.com/albums/m153/vicpret/natgenogps_zps2c1157b8.jpg (http://s103.photobucket.com/user/vicpret/media/natgenogps_zps2c1157b8.jpg.html)

I presume it means anywhere along this road ?

and

http://i103.photobucket.com/albums/m153/vicpret/circles_zpsd1fdb7a1.jpg (http://s103.photobucket.com/user/vicpret/media/circles_zpsd1fdb7a1.jpg.html)

...............

I ran my DnaTribes and came out as island of Mallorca southern tip ( how did it find the land )

but all other reports just stated PASEO due to the negative data after it computed my initially data

ZephyrousMandaru
01-23-2015, 08:57 AM
What interests me about this tool is many of the combinations I choose frequently result in a faint genetic signal towards South Asia.

kenji.aryan
01-23-2015, 10:05 AM
This tool placed me in Maharashtra lol, which is way off the mark.



you need to register here and Upload your raw data here http://chcb.saban-chla.usc.edu/admixCalc/index.php/home then after you're done click on compute. Then it would show you your admixture results.

After that you have to plug the results in here http://chcb.saban-chla.usc.edu/gps/index.php

MonkeyDLuffy
01-23-2015, 10:07 AM
Based on the data that you have entered our tool computed your closest population to the Rajput region.
Calculated Latitude of Origin: 29.844
Calculated Longitude of Origin: 73.4239
The region is marked on the map below as markers

http://i.imgur.com/TFYD1xC.jpg

MitchellSince1893
01-23-2015, 01:10 PM
Understand the Eurogenes K9 is giving bad data compared to Geno 2.0

Could a Geno 2.0 and 23andme customer answer this for me.

How do the following compare:

Eurogenes K9 to loading 23andme raw data to the site?
Geno 2.0 to loading 23andme raw data to the site?

vettor
01-23-2015, 05:50 PM
first you register here and Upload your raw data here http://chcb.saban-chla.usc.edu/admixCalc/index.php/home then after you're done (It takes few minutes but it can faster sometimes) click on compute. Then it would show you your admixture results.

After that you have to plug the results in here http://chcb.saban-chla.usc.edu/gps/index.php

thanks, i followed your advice

it seems very good although I imagine the first test gave me a GREEK origin from preveza ( ancient Epirotes )
http://en.wikipedia.org/wiki/Preveza

and the second a 92.3% north-italian

http://i103.photobucket.com/albums/m153/vicpret/newcalc_zps9b343ff4.jpg (http://s103.photobucket.com/user/vicpret/media/newcalc_zps9b343ff4.jpg.html)

so, is the Epirote the ancient as I have no Greek link for last 300 year ( i know via birth and marriage certificaes )


................

and my son's results are same as mine in regards to preveza , except 5 km further south

and he is 100% north-Italian

Wulf Warrior
01-23-2015, 06:55 PM
I'll give it a go myself, see what happens!

Alpine Hominin
01-23-2015, 07:01 PM
Decided to give this a shot.
My GPS: I came back in the "Alsatian" and put my location mark a little south-west of Stuttgart, Germany.

http://chcb.saban-chla.usc.edu/gps/record.php?jobID=GPS-23012015-064340-3174

This result does make sense, my paternal haplogroup, surname, and appearance fitting the area. I don't know enough about my family origins to say really how accurate the tool is. With my line also being a possible NPE, I'm relying on Genetics more than paper trail genealogy.

My reAdmix came back 100% "Alsatian_0" as well when I used the "Too mixed" option. When I used the 2 populations one, it came back as follows:

# Population Name Latitude Longitude
Ancestral Route 1 Alsatian_0 48.5 7.5
Ancestral Route 2 Luxembourgish_0 49.61 6.13


This program seems positive I hail from the Germany-France border. This lines up with my other results (When British isn't involved). I didn't input any known ancestral links when running my results. I may go back and put Britain as a known ancestry and see what happens.

Wulf Warrior
01-23-2015, 07:12 PM
Awful pictures, sorry!

Based on the data that you have entered our tool computed your closest population to the British region.
Calculated Latitude of Origin: 52.4826
Calculated Longitude of Origin: -1.20031
The region is marked on the map below as markers.

You may wish to take a note of result reference GPS-23012015-071133-4074 to see the analysis in the future.
3563

ReAdmix Tool 1,

3564

Based on your admixture composition, we found the following ancestral populations
# Population Name Latitude Longitude
Ancestral Route 1 Dane_0 56 10
Ancestral Route 2 Belgian_0 58.85 4.35

Alpine Hominin
01-23-2015, 07:19 PM
Based on the data that you have entered our tool computed your closest population to the British region.
Calculated Latitude of Origin: 52.4826
Calculated Longitude of Origin: -1.20031
The region is marked on the map below as markers.

You may wish to take a note of result reference GPS-23012015-071133-4074 to see the analysis in the future.
3563

I'm glad you happened to post your results just now, I was staring at mine and pondering the programs ability to predict English, as well as other programs ability to differentiate French, German, and English. I score high British in Eurogenes calculators but usually Germany and France dominate my scores in other calculators. Using DMXX's method I scored just a touch north of what these results gave, but it was certainly close enough to merit a mention.

AJL
01-23-2015, 08:57 PM
Okay, it seems rather clear that the problem is with K9b. That's good to know.

ZephyrousMandaru
01-23-2015, 09:26 PM
I'd recommend either uploading your 23andMe raw data to admixCalc or use your Geno 2.0 results (if you tested with them). Eurogenes K9b is very noisy.

MitchellSince1893
01-23-2015, 09:45 PM
Here are some images from GPS. Top 2 are me.
3581

3580

What's interesting about the above is location 1 Denmark is close to my mom's result, while location 2, Luxembourg is close to my dad's location

Mom's
3578
3582

Dad's
3579

vettor
01-23-2015, 09:55 PM
My Wife's data uploaded properly, but has a message of Admixture computation failed.

ZephyrousMandaru
01-23-2015, 10:58 PM
Also, if your raw data is failing to upload, you'll have to download a fresh raw data file. That's the problem I was having, I think it's because I was using a raw data file from an earlier build, that's why it wasn't being uploaded.

MitchellSince1893
01-24-2015, 03:56 PM
Updated previous post with additional maps. I'm thinking the route location 2 for my dad in Bulgaria may reflect his known British India ancestry and/or or apparent Romany ancestry.

Here are some images from GPS. Top 2 are me.
3581

3580

What's interesting about the above is location 1 Denmark is close to my mom's result, while location 2, Luxembourg is close to my dad's location

Mom's
3578

3582

Dad's
3579

3583

vettor
01-24-2015, 05:44 PM
Updated previous post with additional maps. I'm thinking the route location 2 for my dad in Bulgaria may reflect his known British India ancestry and/or or apparent Romany ancestry.

I thought these routes from #1 to #2 indicated an origin "anywhere along the road"..................not necessarly arriving at number 2 or even beginning at number 1

all 3 test at this site are seperate and do not link with each other.....my opinion

MitchellSince1893
01-24-2015, 09:47 PM
I thought these routes from #1 to #2 indicated an origin "anywhere along the road"..................not necessarly arriving at number 2 or even beginning at number 1

all 3 test at this site are seperate and do not link with each other.....my opinion

I'm not sure on the specifics but the only thing that would pull my father's that far southeast would be the Indian and/or Romany ancestry. Without these inputs I would expect both points to be in NW Europe.

AJL
01-24-2015, 09:50 PM
Okay, knowing K9b is suspect, I tried the 23andme upload.

GPS gets me the mountains of central mainland Greece, which is farther southeast as compared to any other utility (23andme, Mcdonald's BGA, Family Finder, MyOrigins, etc.), but still better than Sweden.

reAdmix on simple setting gets me Tuscan/Ingushetia, a snaking line from Florence through the heart of the Balkans and northern Turkey to Armenia -- which is not Ingushetia. Better than Finnish/Bermudan, but still farther southeast than usual.

reAdmix V2 no longer seems to be a separate option. reAdmix on complex setting, though, gives me ridiculous results, with high Abkhaz scores no matter what combination of knowns I pick. (I am not sure why this utility is so obsessed with Russian-controlled areas of the Caucasus in my case.) So while K9b seems awful, I would not recommend the complex mode of reAdmix even using the uploaded coordinates when many better-behaving tools are available. It seems to be "Yukaghirgate" all over again but with Abkhaz.

Salkin
01-24-2015, 10:31 PM
Huh. I got it to accept my Geno 2.0 data, which was a bit fiddly as I seem to have needed to change the extension to .txt. Anyhow, with that data, GPS put me just off the southeastern coast of Iceland. reAdmix simple makes me Finnish / Biaka Pygmy, bizarrely. Complex with Swedish + Belgian gives me 100% Swedish.

Caspian
01-26-2015, 08:33 AM
http://i.imgur.com/6Mr6qqK.png
http://i.imgur.com/3HunmMv.png

That calculator calculated me to Iranian Jews location.

My ancestors may have been similar to Iranian Jews without Turkic influence (11% East Asian genes) 1000 years ago so before Turkification. Also my y-dna found at 20% frequency in Persian speaking Jews (a.k.a Jewish Tats or Mountain Jews) in Azerbaijan and Dagestan. They emigrated from southwestern Iran to Azerbaijan and Pontic steppe during Sassanid and Khazar khanate periods.

icebreaker
01-26-2015, 10:13 AM
I get very strange results.

Shaikorth
01-26-2015, 10:20 AM
Huh. I got it to accept my Geno 2.0 data, which was a bit fiddly as I seem to have needed to change the extension to .txt. Anyhow, with that data, GPS put me just off the southeastern coast of Iceland. reAdmix simple makes me Finnish / Biaka Pygmy, bizarrely. Complex with Swedish + Belgian gives me 100% Swedish.

Did you try putting the Geno 2 percentages you posted here
(http://www.anthrogenica.com/showthread.php?3572-Salkin-s-Geno-2-0) into the GPS tool? Add some fractions to get the total up to 1.0.

Salkin
01-26-2015, 10:27 AM
Did you try putting the Geno 2 percentages you posted here
(http://www.anthrogenica.com/showthread.php?3572-Salkin-s-Geno-2-0) into the GPS tool? Add some fractions to get the total up to 1.0.

I hadn't. And wow, that's probably the best result I've had from any BGA tool. It's only like 80ish km or so from where I actually grew up.

reAdmix on simple gave a route from Finland to Iberia. reAdmix on complex, 100% Swedish.

AJL
01-26-2015, 05:16 PM
It seems the Geno2 results are inherently better. In any event I added my mother's data and she allegedly plots in Belgium, which is consistent with all other BGAs, but the mapping is in the North Sea. In double mode she gets Frisian/Bulgarian. In mixed mode her results are still Abkhaz-obsessed. I have however been able to prove that this is a problem with the calculator or sample set itself. If I select only Abkhaz, my mother's results are 10.7% Abkhaz and nothing else. That's some powerful 10.7%!

vettor
01-26-2015, 05:44 PM
It seems the Geno2 results are inherently better. In any event I added my mother's data and she allegedly plots in Belgium, which is consistent with all other BGAs, but the mapping is in the North Sea. In double mode she gets Frisian/Bulgarian. In mixed mode her results are still Abkhaz-obsessed. I have however been able to prove that this is a problem with the calculator or sample set itself. If I select only Abkhaz, my mother's results are 10.7% Abkhaz and nothing else. That's some powerful 10.7%!

I see no issue with Abkhaz............its an ancient area where ancient Greeks mixed with north caucasus people ............I plotted their with, see below

http://i103.photobucket.com/albums/m153/vicpret/233plotts_zps7a22a1cb.jpg (http://s103.photobucket.com/user/vicpret/media/233plotts_zps7a22a1cb.jpg.html)

In regards to geno2 data ( see map again ).........sure it might be better, as per ftdna project I join one person with same subclade who told me he originally was from Giron in Spain ( although the earliest link between us is not before 800 years )


BTW, my natgeno is post#54

AJL
01-26-2015, 06:21 PM
^ I see what you're saying but speculative links with that region is not the problem -- it's the bad math. No single population can have a genetic signature that is 935% stronger than all others.

Kurd
01-27-2015, 04:00 AM
After reading all the other posts, I decided to give it a try by uploading my data, and having the system calculate the components. I got the following:

http://i1377.photobucket.com/albums/ah42/dil88888/GPS-1_zpshccue1go.jpg (http://s1377.photobucket.com/user/dil88888/media/GPS-1_zpshccue1go.jpg.html)

which placed me in the Teheran, Iran area. So I proceeded to the next step and got 2 ancestral routes, one in Lebanon, and the other in India:

http://i1377.photobucket.com/albums/ah42/dil88888/GPS1_zpsbq3h8mrb.jpg (http://s1377.photobucket.com/user/dil88888/media/GPS1_zpsbq3h8mrb.jpg.html)

So I went on to ReAdmix, and plugged in Kurd, and checked off very confident, since that I am confident of my ancestry, leaving the rest blank and got:

http://i1377.photobucket.com/albums/ah42/dil88888/REAdmix_zpspqzwoa7t.jpg (http://s1377.photobucket.com/user/dil88888/media/REAdmix_zpspqzwoa7t.jpg.html)

Next, I left everything the same, but changed the confidence level to not confident, and got this:

http://i1377.photobucket.com/albums/ah42/dil88888/REAdmix%20Not%20so%20confident_zpsqto1czbg.jpg (http://s1377.photobucket.com/user/dil88888/media/REAdmix%20Not%20so%20confident_zpsqto1czbg.jpg.htm l)

I did not find the program very informative, and would rate it as 6/10.

Humanist
01-27-2015, 04:36 AM
After reading all the other posts, I decided to give it a try by uploading my data, and having the system calculate the components. I got the following:

http://i1377.photobucket.com/albums/ah42/dil88888/GPS-1_zpshccue1go.jpg (http://s1377.photobucket.com/user/dil88888/media/GPS-1_zpshccue1go.jpg.html)

Why are they associating Iranian Jews with the coordinates for Tehran? Basically, they are treating Iranian Jews as Iranians? That is simply wrong. Yeah, not a very reliable tool.

Kurd
01-27-2015, 04:55 AM
Why are they associating Iranian Jews with the coordinates for Tehran? Basically, they are treating Iranian Jews as Iranians? That is simply wrong. Yeah, not a very reliable tool.

That was one thing that I found odd, the other was deferring to Lebanon and India as surrogates for Kurds, even when I checked off that I was very confident that my ancestry is Kurd, and left everything else blank. I was being lenient when I gave it a 6/10.

Humanist
01-27-2015, 05:03 AM
That was one thing that I found odd, the other was deferring to Lebanon and India as surrogates for Kurds, even when I checked off that I was very confident that my ancestry is Kurd, and left everything else blank. I was being lenient when I gave it a 6/10.

Yes. All of it sounds very strange.

Kurd
01-27-2015, 05:35 AM
Things did not get any better by using K9b instead of the uploaded raw DNA file. Now it shifted me further east to around Ghazni, Afghanistan.

http://i1377.photobucket.com/albums/ah42/dil88888/GPS%20Tool%20using%20K9b-2_zpsrqyylxu7.jpg (http://s1377.photobucket.com/user/dil88888/media/GPS%20Tool%20using%20K9b-2_zpsrqyylxu7.jpg.html)

With the ReAdmix tool, after checking off Kurd, leaving everything else blank, and checking off unsure, I got even stranger results. It practically split my ancestry up between Yukagir and S. India, with a little bit of Spanish for good measure, and totally ignored Kurd and all other Iranic groups

http://i1377.photobucket.com/albums/ah42/dil88888/Readmix%20Unsure%20k9b_zpst1tebhac.jpg (http://s1377.photobucket.com/user/dil88888/media/Readmix%20Unsure%20k9b_zpst1tebhac.jpg.html)

AJL
01-27-2015, 03:59 PM
Why are they associating Iranian Jews with the coordinates for Tehran? Basically, they are treating Iranian Jews as Iranians? That is simply wrong. Yeah, not a very reliable tool.

Not only that, when I ran an entirely Ashkenazi person through GPS using K9b they were in Belgium. With reAdmix on split mode they ended up being dual Norwegian/Iberian, and on multiple mode with Ashkenazi selected at medium confidence:

Suomalainen_0 Finland 6.6%
Kuwait_2 Kuwait 15.6%
Abkhazian_1 Georgia 0.1%
Yoruba_2 Nigeria 0.3%
Ashkenazy Jews N_0 Netherlands 4.5%
Bulgarian_2 Bulgaria 17%
Yukaghir_3 Russia 3.5%
Bermudian_2 Bermuda 7.2%
Ethiopian_0 Ethiopia 0.1%
Portuguese_0 Portugal 11%
Dane_0 Denmark 15.7%
BiakaPygmy_3 Central African Republic 18.3%

The utility has perhaps made the mistake of locating diaspora populations in host countries. "Ashkenazim in the Netherlands" should not be "Belgium."

vettor
01-27-2015, 06:28 PM
Not only that, when I ran an entirely Ashkenazi person through GPS using K9b they were in Belgium. With reAdmix on split mode they ended up being dual Norwegian/Iberian, and on multiple mode with Ashkenazi selected at medium confidence:

Suomalainen_0 Finland 6.6%
Kuwait_2 Kuwait 15.6%
Abkhazian_1 Georgia 0.1%
Yoruba_2 Nigeria 0.3%
Ashkenazy Jews N_0 Netherlands 4.5%
Bulgarian_2 Bulgaria 17%
Yukaghir_3 Russia 3.5%
Bermudian_2 Bermuda 7.2%
Ethiopian_0 Ethiopia 0.1%
Portuguese_0 Portugal 11%
Dane_0 Denmark 15.7%
BiakaPygmy_3 Central African Republic 18.3%

The utility has perhaps made the mistake of locating diaspora populations in host countries. "Ashkenazim in the Netherlands" should not be "Belgium."

This "belgium" in the ancient times represented modern belgium, luxembourg and netherlands today , is this what they meant?
If they meant modern times, then would not the ashkenazi in nerherlands only happen in majority after 1492 when all faiths where thrown out of spain and portugal except for catholics. The jews of these land settled in netherlands by the bucket full.

I am not defending the program because I believe in it, I simply want to know their reasoning so I can make a true decision on how good this program is. To me the natgeno simplified data exceeds their data, K9b and 23andme .................maybe I should try MyOrigins

vettor
01-27-2015, 06:46 PM
I tried myorigns data and it gave me a point near Lyon frnace....it is possible to be correct

The program has updated overnight and doing the admix2 , one get a different view ..with a map

Iberian_0 Spain 48.8%
Bantu S.E. S.Sotho_0 Lesotho 8%
French (Basque)_0 France 7.3%
Tatar_0 Russia 35.9%

and 4 populations



I reran their program as they have changed format .
Below is the readmix using my data to them ( they must be making changes to the program as people join )

Population/Sub-population Common country for this population/sub-population Ancestral composition
Italian (North)_2 Italy 81.4%
Abkhazian_1 Georgia 6.8%

So above is .....from abkhazian to north-italy ( very plausible )

yugahium and bermudian are now gone from my numbers....BTW the default of Preveza in Greece stayed the same

AJL
01-27-2015, 06:49 PM
If they meant modern times, then would not the ashkenazi in nerherlands only happen in majority after 1492 when all faiths where thrown out of spain and portugal except for catholics. The jews of these land settled in netherlands by the bucket full.


Yes, clearly that happened, but no reputable autosomal tool finds that Ashkenazim are centred near Belgium, any more than colonial Americans should be genetically centred along the Atlantic Seaboard.

ZephyrousMandaru
01-27-2015, 08:49 PM
Strangely enough, when I do choose Iraqi or Iranian Jews, I always end up being 100% Iranian Jewish or Iraqi Jewish. So the problem partly lies in where these populations are being anchored.

Caspian
01-28-2015, 10:42 PM
http://i.imgur.com/6Mr6qqK.png
http://i.imgur.com/3HunmMv.png

That calculator calculated me to Iranian Jews location.

My ancestors may have been similar to Iranian Jews without Turkic influence (11% East Asian genes) 1000 years ago so before Turkification. Also my y-dna found at 20% frequency in Persian speaking Jews (a.k.a Jewish Tats or Mountain Jews) in Azerbaijan and Dagestan. They emigrated from southwestern Iran to Azerbaijan and Pontic steppe during Sassanid and Khazar khanate periods.

+

New results with Geno 2.0 raw data are here. (I used FTDNA raw data for previous calculation)

That calculator calculated me to Azeri speaking population in Northern Iraq a.k.a Iraqi Turkmans. (Iraqi Turkmans speak a dialect of Azeri language)

http://i.imgur.com/cAzUpAj.png
http://i.imgur.com/3snUAwZ.png

And a little part of my MDLP K23B oracle results here.

Using 3 populations approximation:
1 50% Assyrian_Iraqi +25% Turkmen_Uzbekistan +25% Uzbek_Tashkent @ 3.128006

Using 4 populations approximation:

Assyrian_Iraqi + Assyrian_Iraqi + Azeri + Turkmen_Uzbekistan @ 3.217230
Assyrian_Iraqi + Assyrian_Iraqi + Hazara + Kakheti @ 3.356529
Assyrian_Iraqi + Hazara + Iraqi_Chaldean + Kakheti @ 3.381796
Assyrian_Arzni + Assyrian_Iraqi + Kurd + Turkmen_Uzbekistan @ 3.384953
Assyrian_Arzni + Assyrian_Iraqi + Kurd_North + Turkmen_Uzbekistan @ 3.412946

I have some Iraqi and Kurd mixture according to K23B and these results might be good example for Iraqi Turkmans.

Stephen1986
01-29-2015, 12:01 AM
I tested both mine and my brother's 23andMe raw data. We are mostly of English ancestry, with some Irish and Welsh, possible Scottish ancestry and two of our maternal great great grandfathers are not known.

My results were as follows -

Based on the data that you have entered our tool computed your closest population to the AngloSaxon region.

Calculated Latitude of Origin: 54.1563
Calculated Longitude of Origin: -0.722622

My results using my Geno 2 percentages (45% Northern European, 37% Mediterranean and 17% Southwest Asian and adding an extra .333 to each one) -


Based on the data that you have entered our tool computed your closest population to the Dane region.
Calculated Latitude of Origin: 57.2668
Calculated Longitude of Origin: 9.06283
The region is marked on the map below as markers.

My brother's results -

Based on the data that you have entered our tool computed your closest population to the Bavarian region.

Calculated Latitude of Origin: 49.5947
Calculated Longitude of Origin: 9.60339

This isn't a surprising result as my brother tends to show more of the supposed continental ancestry that various genetic analyses say we have.

firemonkey
03-05-2015, 11:18 AM
For K=9
Component 1
0.006941
Component 2
0.377743
Component 3
0.036906
Component 4
0.188921
Component 5
0.050700
Component 6
0.003088
Component 7
0.017690
Component 8
0.318001
Component 9
0.000010


copied and pasta data in order given in my previous post to readmix gps tool and got


http://i4.photobucket.com/albums/y126/labile/gps%20calculation.png

http://i4.photobucket.com/albums/y126/labile/gps%20map.png


My known ancestry is English,Scottish, and Irish!! I know I was born in the Belgium Congo (my father was a junior British diplomat then) but it's patently ridiculous!!


I then did the tool where you can choose several populations and chose British , Irish, Scottish(known ancestry) Welsh( have ancestor born in Wales but not sure if Welsh) Scandinavian(30% according to My origins) and lastly the BiakaPygmy that the GPS tool had given.


This gave http://i4.photobucket.com/albums/y126/labile/ancestral%20composition.png

BiakaPygmy_3 Central African Republic 79.9%

firemonkey
03-05-2015, 07:55 PM
Using ftdna autosomal instead of geno 2 was far more accurate.

http://i4.photobucket.com/albums/y126/labile/gps%20calculation%20ftdna.png

http://i4.photobucket.com/albums/y126/labile/gps%20map%20ftdna.png

paulgill
04-10-2015, 07:59 AM
Just an error, I guess.

kenji.aryan
05-17-2015, 09:38 AM
Here is mine for two pops readmix 2 DON'T QUOTE .

#1 Nattukottai chettiar
#2 Pathan[/I]

fil
05-18-2015, 12:48 AM
i keep getting Admixture computation failed.

Theconqueror
05-31-2016, 07:13 PM
I just saw this thread and I tried it quickly. Here are my results: my closest population location is in Sjaelland (Denmark), as part of what GPSP called the Finnish region.

9548

9549