PDA

View Full Version : Different ancestry composition from FTDNA, 23andMe and NG Geno 2.0



Petr
07-08-2014, 07:35 AM
Does anybody has an idea why there is so big difference in ancestry composition determined by various testing companies? My results are the following:

23andMe, standard (speculative):
25.9% Eastern European (36,8%)
18.5% Northern European (34,6%)
2.3 % Southern European (7,4%)
53.1% Broadly European (21,2%)
-----------------------
99.8 % European

FTDNA, MyOrigins:
56% Eastern Europe
36% Scandinavia
7% Southern Europe
----------------------
99% European

National Geographic - The Genographic project:
Northern European: 47%
Mediterranean: 34%
Southwest Asian: 17%


Or my friend:
23andMe, standard (speculative):
4.3% Eastern European (14.7%)
11.8% Northern European (19.2%)
3.9 % Southern European (12.9%)
44.0% Broadly European (15.6%)
34.6% Ashkenazi (36.6%)
-----------------------
98.5 % European (98.9%)

FTDNA, MyOrigins:
29% Eastern Europe
18% British Isles
6% Scandinavia
46% Ashkenazi Diaspora
1% Eastern Middle East

It seems that 23andMe, FTDNA and NG use different methodology, is there any comparison of the method they use?

Petr

AJL
07-08-2014, 01:40 PM
Hello Petr,

Welcome to Anthrogenica. Each of the companies' analyses varies slightly partly because they all define the regions/groups a little differently, partly because they use different combinations of SNPs with different algorithms, and partly because they
have different customer databases. I suspect none of the databases have many Central Europeans such as Czechs, Austrians, Hungarians, etc., and Genographic is probably the crudest in terms of analysis.

For a (likely better) read on your and your friend's ancestry than is offered by any of the companies, you could try uploading results to Gedmatch.com and running the ancestry analysis tools with the oracles.

There are also several threads on the meaning of the different populations specific to the companies -- for examples here:

http://www.anthrogenica.com/showthread.php?517-23andMe-Ancestry-Composition-Results&highlight=location

http://www.anthrogenica.com/showthread.php?2398-New-Population-Finder-Update-Finally

Alpine Hominin
07-08-2014, 05:37 PM
Welcome to Anthrogenica. As AJL said, the use of different base populations leads to some oddities. If one test includes a separate East, West, North, and south Europe, but another only north and south, you will receive two different results. Gedmatch.com is my favorite collection of calculators, but the site is currently down because they're changing server locations. It's worth the wait though.

vettor
07-08-2014, 11:29 PM
Does anybody has an idea why there is so big difference in ancestry composition determined by various testing companies? My results are the following:

23andMe, standard (speculative):
25.9% Eastern European (36,8%)
18.5% Northern European (34,6%)
2.3 % Southern European (7,4%)
53.1% Broadly European (21,2%)
-----------------------
99.8 % European

FTDNA, MyOrigins:
56% Eastern Europe
36% Scandinavia
7% Southern Europe
----------------------
99% European

National Geographic - The Genographic project:
Northern European: 47%
Mediterranean: 34%
Southwest Asian: 17%


Or my friend:
23andMe, standard (speculative):
4.3% Eastern European (14.7%)
11.8% Northern European (19.2%)
3.9 % Southern European (12.9%)
44.0% Broadly European (15.6%)
34.6% Ashkenazi (36.6%)
-----------------------
98.5 % European (98.9%)

FTDNA, MyOrigins:
29% Eastern Europe
18% British Isles
6% Scandinavia
46% Ashkenazi Diaspora
1% Eastern Middle East

It seems that 23andMe, FTDNA and NG use different methodology, is there any comparison of the method they use?

Petr

why the difference is because I imagine all these companies do not want to use the title Central European, so some companies put swiss with french, austrians with germans, slovenes with italians, czech with german, slovak with poland etc

I have never seen central european designated in any testing company

geebee
07-09-2014, 12:23 PM
Ancestry DNA originally did use "Central European". They had me as 48% Central European, 23% Scandinavian, 16% Eastern European, 6% Finnish/Volga-Ural, and 7% uncertain.

Their current estimate utilizes many more populations and has a lot more specificity.

Now that they are using more populations and identifying 1% of my DNA as Native American -- plus 3% as "West Asian" (Caucasus) and another 3% as "Asian" (Asian Central and Asian South) -- my overall European ancestry has moved somewhat further west.

In Europe, I have "Europe West" (44%), "Great Britain" (27%), "Iberian Peninsula" (9%), Ireland (7%), "Italy/Greece" (2%), "Scandinavia" (2%), "European Jewish" (<1%), and "Finland/Northwest Russia" (<1%). This is much closer to my paper trail ancestry, which is roughly half German, a quarter British (mostly Scottish and Scots-Irish), and an eighth Menorcan Spanish. The remaining eighth includes some other places in Europe, plus some Native American ancestry (probably, but not definitely, Choctaw).

Mark D
07-11-2014, 01:05 PM
An excellent comparison was posted by Roberta Estes on her blog back in May; http://dna-explained.com/2014/05/19/ethnicity-percentages-second-generation-report-card/

Once GedMatch comes back online, you might notice that even Dienekes' several Dodecad programs report entirely different percentages depending on the particular "ethnic" data used. Most of these and several from the companies use public databases from older academic projects such as Human Genome Diversity Project (HGDP) and the more recent 1000 Genomes Project. There are many issues associated with the public databases. 23andMe and now FTDNA add data from their own private databases.

I've tested with FTDNA, 23andMe, Ancestry.com and Geno 2.0. They are consistent only in the most general way, (I'm of European descent) and differ considerably in the details. I take them with a grain of salt and look forward to the advancement of the science.

AppalachianGumbo
07-24-2014, 08:29 PM
An excellent comparison was posted by Roberta Estes on her blog back in May; http://dna-explained.com/2014/05/19/ethnicity-percentages-second-generation-report-card/

Once GedMatch comes back online, you might notice that even Dienekes' several Dodecad programs report entirely different percentages depending on the particular "ethnic" data used. Most of these and several from the companies use public databases from older academic projects such as Human Genome Diversity Project (HGDP) and the more recent 1000 Genomes Project. There are many issues associated with the public databases. 23andMe and now FTDNA add data from their own private databases.

I've tested with FTDNA, 23andMe, Ancestry.com and Geno 2.0. They are consistent only in the most general way, (I'm of European descent) and differ considerably in the details. I take them with a grain of salt and look forward to the advancement of the science.

This is true. Also, such groups as Native Americans are not often not included in private databases from clients that personally tested with DTC as their European, African (not New World Americas) clients are. As you mentioned, some statistics are gathered from older projects. Nat. Amer. in the HDGP was from diabetic research but only existing of small populations within Central/South Indigenous Americans, slim pickins.

With 23andMe for example, 5k Europeans an 108 Nat. Amer., smaller Pacific Islanders makes for very sub-optimal analysis of certain geographcial reigions.