DNA Results Differ Between 23andMe, MyHeritage, and GEDmatch: Which is correct?

Question

Well, I've tested with 23andMe as well as MyHeritage and I've come back with conflicting reports from both. But first things first, I've never really known my heritage. Not a lot of it. As far as I'd known prior to taking the test on my father's side I'm: Irish, British, Scottish, Native American, with smaller amounts of Welsh, French, Italian, Dutch, and Danish. And my mother's side is: Mexican, Native American, Irish, Scottish, British, French, German, with smaller amounts of Welsh. But there are a lot of people I'm not sure what they were as far as nationalities and ethnicities. So when I took my DNA test, it was a bit of a shock to say the least. Here's my results.

23andMe Results:
British & Irish - 48.5%
French & German - 3.6%
Scandinavian - 2.3%
Finnish - 0.3%
Broadly Northwestern European - 26.1%
Iberian - 4.5%
Italian - 0.4%
Broadly Southern European - 3.5%
Broadly European - 4.1%
Native American - 4.9%
East Asian - < 0.1%
Broadly East Asian - < 0.1%
Broadly East Asian & Native American - 0.8%
West African - 0.5%
Broadly Sub-Saharan African - 0.1%
North African - <0.1%
Unassigned - 0.3%

MyHeritage Results:

English - 69.5%

Italian - 8.3%

East European - 5.0%

Central American - 12.2%

West Asian - 5.0%

GEDmatch Results: Eurogenes K13

Admix Results (sorted):

# Population Percent
1 North_Atlantic 43.18
2 Baltic 19.88
3 West_Med 11.61
4 Amerindian 8.57
5 East_Med 7.45
6 West_Asian 7.14
Using 1 population approximation:
1 West_German @ 9.097307
2 South_Dutch @ 9.878935
3 Southeast_English @ 12.224097
4 North_German @ 12.362290
5 French @ 12.436082
6 Orcadian @ 13.412692
7 Danish @ 13.484862
8 North_Dutch @ 13.504702
9 Southwest_English @ 13.950716
10 Irish @ 14.006430
11 West_Scottish @ 14.494095
12 Austrian @ 14.780554
13 Norwegian @ 15.400504
14 Swedish @ 16.279316
15 East_German @ 16.317949
16 Hungarian @ 18.058197
17 Spanish_Cataluna @ 19.684835
18 North_Swedish @ 20.363480
19 Portuguese @ 20.506191
20 Spanish_Castilla_Y_Leon @ 20.677061

Using 2 populations approximation:
1 50% West_German +50% West_German @ 9.097307

Using 3 populations approximation:
1 50% Irish +25% Italian_Abruzzo +25% Norwegian @ 8.562629
Using 4 populations approximation:
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1 Irish + Irish + Italian_Abruzzo + Norwegian @ 8.562629
2 Irish + Italian_Abruzzo + Norwegian + Orcadian @ 8.573209
3 Ashkenazi + Irish + Irish + Irish @ 8.592313
4 Irish + Italian_Abruzzo + Norwegian + West_Scottish @ 8.594319
5 Norwegian + Norwegian + South_Italian + West_Scottish @ 8.599518
6 Irish + Norwegian + Norwegian + South_Italian @ 8.605319
7 Italian_Abruzzo + Norwegian + Orcadian + West_Scottish @ 8.610585
8 Irish + Irish + Norwegian + South_Italian @ 8.617249
9 East_Sicilian + Irish + Irish + Norwegian @ 8.618892
10 Italian_Abruzzo + Norwegian + Orcadian + Orcadian @ 8.619877
11 East_Sicilian + Irish + Norwegian + West_Scottish @ 8.620096
12 Irish + Irish + Italian_Abruzzo + Swedish @ 8.620117
13 Irish + Italian_Abruzzo + Swedish + West_Scottish @ 8.627582
14 Irish + Irish + Norwegian + West_Sicilian @ 8.628523
15 Irish + Norwegian + South_Italian + West_Scottish @ 8.631497
16 Ashkenazi + Irish + Irish + Norwegian @ 8.631778
17 Irish + Italian_Abruzzo + Norwegian + Norwegian @ 8.638668
18 Central_Greek + Irish + Norwegian + West_Scottish @ 8.639158
19 East_Sicilian + Norwegian + West_Scottish + West_Scottish @ 8.641149
20 Ashkenazi + Irish + Irish + West_Scottish @ 8.643385

Gedrosia K12

Admix Results (sorted):
 

#	Population	Percent
1	SINTASHTA_STEPPE_HERDERS	34.62
2	EARLY_EUROPEAN_FARMERS	31.86
3	CAUCASUS	17.34
4	SW_ASIAN	4.86
5	E_SIBERIAN	3.63
6	BALOCHI	3.58
7	SE_ASIAN	1.69
8	S_INDIAN	1.18

Finished reading population data. 87 populations found.
12 components mode.

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

Least-squares method.

Using 1 population approximation:
1 Greek @ 20.300722
2 Norwegian @ 21.912773
3 Sicilian @ 27.431810
4 Russian @ 30.601942
5 Turks_Istanbul @ 31.495359
6 Turks_Aydin @ 35.169521
7 Turks_Balikesir @ 35.311371
8 Finnish @ 35.525394
9 Estonian @ 38.223743
10 Turkmen_Afghan @ 39.229286
11 Lithuanian @ 39.606781
12 Uzbek @ 39.752441
13 Tajik_Pomiri @ 41.288673
14 Kurds_C @ 41.566750
15 Kurds_N @ 41.575188
16 Turks_Adana @ 42.033325
17 Turks_Kayseri @ 42.687382
18 Kurds_F @ 43.851204
19 Iraqi_Chaldeans @ 44.079502
20 Iraqi_Arab_Baghdad @ 44.318951

Using 2 populations approximation:
1 50% Norwegian +50% Turks_Istanbul @ 6.315009


Using 3 populations approximation:
1 50% Norwegian +25% Sicilian +25% Turks_Istanbul @ 5.706852


Using 4 populations approximation:
+++++++++++++++++++++++++++++++++++++++
1 Greek + Norwegian + Norwegian + Turks_Istanbul @ 4.692763
2 Greek + Norwegian + Norwegian + Turks_Aydin @ 4.738549
3 Greek + Norwegian + Russian + Turks_Istanbul @ 4.749650
4 Greek + Norwegian + Norwegian + Turks_Balikesir @ 4.875467
5 Finnish + Norwegian + Sicilian + Turks_Istanbul @ 4.962596
6 Norwegian + Russian + Sicilian + Turks_Istanbul @ 4.989967
7 Estonian + Norwegian + Sicilian + Turks_Balikesir @ 5.015533
8 Finnish + Greek + Norwegian + Turks_Istanbul @ 5.053809
9 Estonian + Norwegian + Sicilian + Turks_Istanbul @ 5.181561
10 Finnish + Norwegian + Sicilian + Turks_Balikesir @ 5.182767
11 Lithuanian + Norwegian + Sicilian + Turks_Balikesir @ 5.211343
12 Greek + Norwegian + Russian + Turks_Balikesir @ 5.296608
13 Norwegian + Russian + Sicilian + Turks_Balikesir @ 5.364137
14 Lithuanian + Norwegian + Sicilian + Turks_Istanbul @ 5.413669
15 Finnish + Greek + Norwegian + Turks_Balikesir @ 5.417234
16 Greek + Norwegian + Russian + Turks_Aydin @ 5.443681
17 Estonian + Greek + Norwegian + Turks_Balikesir @ 5.458034
18 Estonian + Greek + Norwegian + Turks_Istanbul @ 5.470024
19 Finnish + Greek + Norwegian + Turks_Aydin @ 5.544076
20 Estonian + Greek + Norwegian + Turks_Aydin @ 5.553519

23andMe picks up no West Asian ancestry whatsoever but MyHeritage says 5% while GEDmatch seems insistent its about 7%. So why the differences? Is it actually there or just false readings?

Comments

This is a great question Sharon, one many other people have as well.

Thanks for asking :)

Answer 1

You'll get a number of answers, so I'm just starting the ball rolling.  But I think you'll understand the problem better, by the diversity of these answers.  Similarly, both the similarities and the differences in the admixture reports together produce a clearer picture of your true heritage.

I'm sure you won't like the first part of my answer, because it seems so wrong, but the fact is that while the admixture reports look so different and conflicting, to a large extent they are actually ALL correct.  The problem is they all use different methodologies.  And the science is still very young, constantly changing, improving as they continue gathering data.  Some of the factors that make them so different:

- they classify populations differently

- they include very different populations, both as to region and as to age

- they base their determinations on differing sets of SNP's and genes

- they base their statistics on differing testing groups (often many more English speaking than non-English, especially British Isles, and many more Western civilization than Eastern)

- and there's no way for each geneticist to exclude their own bias's; they each tend to specialize on populations of specific regions and time periods

It's important also to understand what general population age each analysis is based on.  For example, if populations like German, Italian, and French are mentioned, then you know the analysis and report are concentrating on very recent times, within the last 2 millennia only.  If you see hunters and gatherers mentioned, then you know they are talking about much older populations, by many thousands of years.  The reports from these 2 examples would be totally different, and yet both would be correct.  I really really wish that all admixture reports would provide a general time period for the populations they associate with your results.  Users need to know that while we came from certain countries in recent times, those people came from very different regions thousands of years before that.  For example, take the Amerindian population.  If you find evidence of Amerindian, then you *have* to find evidence of East Asians too, because Native Americans came from East Asia.  But then you should also be able to find West Asian, because East Asians came from West Asia!

There have been so many human migrations, waves of populations flowing one way, then back again, and later repeating a similar trail, and mixing with the previous migrations.

This is why I don't believe you should ever trust the results of any one analysis.  You will always get a better picture by combining multiple reports.  And those exact percentages they like to show?  Ridiculous!  5% is the same as 7%, and anything less than 1% is unreliable.

Comments

Thank you for your explanation Rob; it definitely helps me see things more clearly.

---

An excellent answer, but to add a little to it, you also need to look from your own perspective backwards. You did not receive an equal percentage of your DNA from each of your ancestors, go back even a few generations and there will be some of you GGGG grandparents from whom you inherited large tracts of DNA and others from whom you have inherited none at all.

You could have had a GGGG grandfather who was the only Chinese member of the family and through the roulette wheel of inheritance got absolutely no fragments of DNA from him - your Chinese heritage would just not show in any test irrespective of what database or algorithm they use.

---

Hm. So basically what you're saying is that all of my results from GEDmatch are true then? Wow. I guess I kind of had started to believe what I'd been reading - since most people just say that not all of the results it gives are true. But I don't see why it would come up if it's not indeed there. It makes sense that it would be there, otherwise the test wouldnt' say so.

---

My Wife, and her Mom and Dad all took 23andMe tests.  I agree with the first commentor, that all the ethnic mixes were "About Right". And the DNA confirmed that there was no "NPE" or "Switched at Birth" issues.  However, we were kind of expecting that my wife would be a more or less "Perfect Average" between her parents.  Not So.  In fact my wife shows up as something tlike 92% British Isles, her dad 90% and her mom 85%.  Not the kind of math I was expecting.

But then I was told that markers were chosen at random from each persons DNA (and not necessarily the same markers) and compared to something like 60,000 people of "High Confidence" in their genealogical lineage.  Different random samples will produce different resuts.

I am a Software engineer, and I do a fair amount of work with Encryption. The idea behind Encryption is to create a complex math problem which requires exact inputs to reverse the encryption. Currently an encryption key made from 128 bits of information is considered highly secure (128 unique pieces of information that could individually be either a 0 or a 1). The only way to "Solve the Problem" when you don't know the key, is to individually try each possible combination, and 2 to the power or 128 is a *very* big number that would take modern computers many centuries to solve by "Guess and Check".

DNA has *way* more unique pieces of information.  Even when considering Sequences instead of unique bits. To fully compare the genome of even a single person to a known database of 60,000 would take many computers many lifetimes.  Therefore, they take a Random Sampling, to get a close approximation, but for the foreseeable future there will probably never be a margin of error on the numbers less than 3-5% or so.  It's like political polling, and the saying: "The only poll that counts is the one on election day".

---

> "So basically what you're saying is that all of my results from GEDmatch are true then?"

Sharon, I'm not comfortable using the word 'true' for any prediction (or even 'not true').  'Estimate' would be a much better word, because that's what they are, with sizable margins of error.  Each test is very different, and the rough estimates they present have to be taken with a large grain of salt, keeping in mind as best you can what the objectives and database for that test are.  For example, a test with little Amerindian input is not likely to be at all reliable with Native American predictions.  Tests with strong British Isles participation are likely to be more accurate with British Isles predictions/estimates.

---

Three more discussions and articles on this subject, all very helpful:

- A Gedmatch Admixture Guide!

- How valid is family tree DNA "origins"?

- Those percentages, revisited

---

I love your answer, Rob. I think this is the clearest, non-technical explanation I've seen on any site.

---

Another way of looking at it is Blaine Bettinger’s atDNA chart used by ISOGG, of cM range by cousin distance. It shows even at third cousin level the range is 0-217 cM, meaning that in as little as 4 generations Derrick’s potential Chinese ancestor has gone forever.

A friend said she had Jewish ancestry, but she may not have. What she had was a polymorphism that is much more common in Jewish people than it is in the general population because of marriage preferences.

As Brian indicates, many grains of salt are required when looking at these estimates. I’m happy being British, infused by wave after wave of people crossing from Europe.

Answer 2

West Asian, West Mediterranean, East Mediterranean can all go together if you are Italian.  I am Italian and I get all three.  Greek and West Asian are common for Italians to get in their DNA tests.  So it's possible this is what those are in yours.  Also the "Broadly Southern European" could be Italian or could be Spanish (Iberian).

Can you run a Eurogeno K36 test on GEDmatch and check the amount of your Iberian though?  Check the amount of Iberian vs the amount of Italian on that test should tell you something.

The Siberian and East Asian make sense for Amerindian since they have discovered Amerindians originated there.
 

I really don't see any discrepancies on your tests except the first one didn't show much Amerindian.  But the focus of the tests are different and 23andme uses the "Broadly xxx" category because they don't have enough samples in their database for many of us.  For example Swiss people get "Broadly European" or "French/German."  In fact, Northern Italians can also get "French/German" from 23andme.  It doesn't mean discrepancies just that that's the only category they have for Swiss, whereas GEDmatch can specifically specify Swiss or even North Italian.

I kept wondering where the West Asia was coming from in mine and it's always 5-7% I'm thinking part of Greece by Turkey, still looking.  I discovered many Italians have Greek DNA and unfortunately will be confusing because there's no telling what category GEDmatch or 23andme will put it.  So one test will list my Greek as Eastern Mediterranean and on another as Southern European.  And 23andme they just lumped it into "Broadly Southern European.

 

=========================================

Ref:

East Med 13.23

Cyprus, Greece,  Lebanon, Syria, Palestine, Israel, Turkey, Egypt, Jordan. The more broad definition of the Levant which includes its historically tied neighboring countries, Greece and Egypt.  The region of Syria with the island of Cyprus (also known as the Levant), and Turkey, which limits the definition to Western Asia.

West_Asian, 6.51

The following countries and areas comprise the West Asia subregion: Afghanistan, Armenia, Azerbaijan, Bahrain, Cyprus, Gaza Strip, Georgia, Iran (Islamic Republic of), Iraq, Israel, Jordan, Kuwait, Lebanon, Oman, Qatar, Saudi Arabia, Syrian Arab Republic, Turkey, United Arab Emirates, West Bank and Yemen

Answer 3

Sharon Each of those Testing Companies come up with those admixture Results based on your Autosomal Matches in there database, that is why the results vary from company to Company.
The Only way you will get an accurate admixture is Full Genome Testing. For example MT Full Sequence and have your Brother or Father do Y111 and Big Y

Answer 4

I have been told several times that the ethnicity or origins is an ESTIMATE or a best GUESS. That is all they are. They are not at all accurate.

I was inclined to think that they were accurate until I did my own test.

The FT DNA said that I was 22% EAST European - but NONE of my matches reflected that.

MY heritage says I am 22% Scandinavian, and since I know I have ancestors from Orkney, I am much more inclined to accept tthat estimate.

MY heritage also said I was Scottish/Irish/Welsh but not English. FT DNA just says I am british. My top matches are showing an irish family.

Answer 5

The short answer is that the ethnicity percentages the various companies give you are hopelessly inaccurate, and always will be. You can count on their telling you what continent your ancestors came from, which is useful only if you look in the mirror and can't tell if you're a white guy, a black guy, or an asian guy, but that's about it. They exist entirely to offer you an "instant genealogy" - it's a sales gimmick. It's practically fraud.

It is no substitute for real genealogical research, and it's real value is the DNA matches, which can reveal and/or confirm your fairly recent ancestry (within 5 generations or so). They don't want to actually advertise that, though, because that requires some education of and work by the customer, and tells them that their results may reveal something that they may not be prepared to know.

Comments

Note that this is a question from 2017.

---

I saw that, but it's irrelevant. The same basic question will keep coming up - it's timeless. People who have the same basic question, but haven't bothered to ask, will come across this old question and will see what others think about it. The answer hasn't changed either.