Pop Finder results vs other tests - how do yours compare?

[S9arthur]

I wrote to FTDNA asking about the interpretation of my Population Finder results, but haven't received a response yet, so I thought I would bring it up for discussion here.

I have tested with multiple companies, and have received various results on my ancestral makeup. On Ancestry by DNA's DNA Print 2.5, I was 86% European, 8% Native American, and 6% Sub-Saharan African. On DecodeMe, my results were 92% European, 4% Asian, and 4% African. Obviously DNA Print was mistaking the "East Asian" for Native American, a problem that several tests have apparently suffered from. On 23andMe, I was 99% European, and <1% each of Asian and African. This was recently revised by them to about 99% European and about 1% African. I also sent my 23andMe raw data to Dr Doug McDonald, who came up with the same numbers (99% Euro, 1% Afr).

Traditional, or "paper" genealogy suggests perhaps 1% - 5% African (or more precisely "Atlantic Creole," a mix of Angolan and Portuguese), the rest mostly English, with some Irish, French, Swedish, and possibly German. Although several of my ancestral families have the ubiquitous "Cherokee Princess Grandmother" legends, I never put a lot of stock in these stories, as they are fairly easily dismissed using traditional genealogical and historical methods.

Population Finder, however, picks up no African, only about 95% Orcadian and about 5% Middle Eastern. I'm not sure what to make of this. Is the African being picked up by the other tests an error, or is there some flaw with PF? If I am a mix of mostly English with small amounts of African and so on, would that "pull" my geographic estimate southeastwards, so to speak? Or does PF have problems with small amounts of ancestry, and thus simply can't pick it up?

Has anyone else tested with multiple services and received varying results?

[S9arthur]

Almost as soon as I posted the above, I received the following reply from FTDNA:

"Hello,

Thank you for your email. We do not report any ethnicity that is less than about 3% because of the potential margin of error. Please also note, the Population Finder program is currently in the Beta phase. Small adjustments may need to be made as the Bioinformatics team responds to user feedback."

I wrote back suggesting they post a note about not reporting ancestries below 3% on the customer's PF results page.

[Daniel72]

Well I have tested with 2 companies:

FTDNA and 23andMe

I also have send my aDNA raw data to several DNA Projects that analysed them or keep on testing and analysing them with various technologies, while at the same time trying to find new ways of analysing DNA.

Most of my results are in my sig.

FTDNA
Affy:
85% Western European
15% European

New Chip, first try:
91% Orcadian, French, Spanish
9% Finnish, Tuscan

New Chip second try:
92% Orcadian
8% Middle Eastern

23andMe:
>99% European, <1% Asian, 0% African
23andMe Anchestry Finder: 1. German, 2. British, 3. Norwegian, 4. French

DrMcDonalds
Analysis of 23andMe Raw Data:
98.5% English, 1.5% Asian, impossible to say what, because to little fragments.

Eurogenes Project
Analysis of 23andMe Raw Data:
Absolutely German, nowhere "in between".
But also: All Eurogenes calculations repeatetly show higher than average German levels of Finnish, Baltic, Chuvash or NE Asian etc.

Dodecad Project
Various experiments on 23andMe Raw Data:
a) Celto-Germanic or Balto-Slavic? ---> Celto-Germanic

b) Clusteres with CEU, Irish, British, Scandinavians and Germans

c) Clusteres with Scandinavians and Germans

d) Clusteres with 60% of the Germans, 50% of the Austrians, 50% of the Hungarians and all (2) Slovenians into a "Central European" cluster.

I meanwhile also joined the "Artemis Project", but no results so far.

[nathanm]

I wrote to FTDNA asking about the interpretation of my Population Finder results, but haven't received a response yet, so I thought I would bring it up for discussion here.

I have tested with multiple companies, and have received various results on my ancestral makeup. On Ancestry by DNA's DNA Print 2.5, I was 86% European, 8% Native American, and 6% Sub-Saharan African. On DecodeMe, my results were 92% European, 4% Asian, and 4% African. Obviously DNA Print was mistaking the "East Asian" for Native American, a problem that several tests have apparently suffered from.

That's not a mistake, it's the literal truth that Native American and East Asian DNA are closer related than to African or European. The results from any of these tests depend on what they're using for reference populations. And the science behind the tests is still very new.

On 23andMe, I was 99% European, and <1% each of Asian and African. This was recently revised by them to about 99% European and about 1% African. I also sent my 23andMe raw data to Dr Doug McDonald, who came up with the same numbers (99% Euro, 1% Afr).

Traditional, or "paper" genealogy suggests perhaps 1% - 5% African (or more precisely "Atlantic Creole," a mix of Angolan and Portuguese), the rest mostly English, with some Irish, French, Swedish, and possibly German. Although several of my ancestral families have the ubiquitous "Cherokee Princess Grandmother" legends, I never put a lot of stock in these stories, as they are fairly easily dismissed using traditional genealogical and historical methods.

Population Finder, however, picks up no African, only about 95% Orcadian and about 5% Middle Eastern. I'm not sure what to make of this. Is the African being picked up by the other tests an error, or is there some flaw with PF? If I am a mix of mostly English with small amounts of African and so on, would that "pull" my geographic estimate southeastwards, so to speak? Or does PF have problems with small amounts of ancestry, and thus simply can't pick it up?

Has anyone else tested with multiple services and received varying results?

"Flawed" isn't the right word. This field is still in its infancy. Both the science and technology, separately, are getting better. FTDNA just switched the entire platform they're using for autosomal tests, so there's bound to be growing pains. And eventually I expect they'll use better reference populations. But as any business, they're being conservative with their estimates. The hobbyist projects can afford to throw caution to the wind and try out the new reference populations. It's not like people are paying them for the analysis.

How do you know the "Atlantic Creole" mix is specifically Angolan? Because many north and east African populations are heavily admixed with Middle Eastern DNA. And Portuguese is often admixed with Middle Eastern.

My own experience is varied, but reasonable. FTDNA lists me as 93.72% Orcadian, and 6.28% South Asian. The latter was a bit surprising, until I read the FAQ, which suggests Roma ancestry. I don't know the ethnicity of one of my great-great-grandfathers, who theoretically contributed about 6.125% of my DNA. Dr McDonald's first attempt had me as 90% Irish and the rest Mideast. His second try said 56% Orkney, 26% Russian, and 19% Romanian.

On 23andMe, my dad is >99% European and <1% Asian. Dr McDonald got 93.6% Western Europe and 6.4% Mideast. I also just got my dad into the Eurogenes project, and he fits right in with the Polish samples.

[Kasandra]

I don't find PF here to be particularly useful or accurate. I'm currently at 93% Orcadian (UK Ireland) and 7% Middle Eastern. Dr. McDonald said I have a very small piece of ME DNA, small enough to be ignored and the rest is from all over northern Europe and the Orcadian area with my 'spot on the map' being Belgium. Eurogenes the second run in Feb seems perfect to me. 75% French and Irish, 20% North Euro and 4% Baltic. My maternal side is half French and half Irish and my father's side is putatively German but going by family finder matches the German is mostly English and baltic/North Euro (north German, north Polish/Prussian, Norway, Lithuanian, and Finn). I even have a ton of Russian matches at the 6-7 cM level on HIR search. And 1% Middle Eastern. I just think FTDNA's analysis is goofy at the moment.

[S9arthur]

nathanm wrote:

That's not a mistake, it's the literal truth that Native American and East Asian DNA are closer related than to African or European.

I realize that, but it IS a mistake if it tells me I'm Native American when I'm not. Especially when I don't have East Asian either.

And:

The hobbyist projects can afford to throw caution to the wind and try out the new reference populations. It's not like people are paying them for the analysis.

Ancestry by DNA, DecodeMe and 23andMe are all commercial activities, just like FTDNA. Dr McDonald is at the University of Illinois, and I don't think "hobbyist" fits in his case. One hobbyist is the physicist Ken Nordtvedt, whose findings on Y-Haplogroup I have been incorporated into FTDNA's Y-tree.

As it stands right now, the most accurate autosomal test results for me (based on matching my estimates from traditional genealogy) is 23andMe. However, FTDNA is still the best when it comes to the Y chromosome, as they even caught and explained a mistaken SNP reading by 23andMe.

Overall, I would recommend that for everyone who can afford it to test with multiple companies, as they all have certain strengths and weaknesses.

[jah]

ok, so the various analyses differ but what about databases?

At this point, who has the largest database of individuals analyzed for autosomal ancestry--23andMe? FTDNA? DNA Tribes?

Obviously, we're talking both SNPs and STRs here, and wildly varying numbers of each tested (DNA Tribes 27 STRs vs. Illumina chip 600k SNPs for both FTDNA and 23andMe), but number of individuals tested matters.

The reality is that--as has previously stated in this and other threads--each analysis has its relative strengths & weaknesses. But so much basic science remains to be discovered (Epigenetics, anyone?) that even when a person's entire genome can be sequenced for under US 1000, what will still matter is:

who you match in the database

[S9arthur]

I wrote:

As it stands right now, the most accurate autosomal test results for me (based on matching my estimates from traditional genealogy) is 23andMe.

I think I have to take that back. It dawned on me that I had been engaging in some circular reasoning. I had been taking the small African amount that shows up with 23andMe's Ancestry Painting and using it as evidence for reading connections into the traditional genealogical records that really weren't there. After taking a fresh look, I think FTDNA's caution in not presenting ancestry amounts under 3% is probably the wiser approach.

The Middle Eastern component never bothered me as much as it seems to some people here - clearly it's a function of ancient connections between Europe and the Middle East, due to the spread of farming, etc.