Why is there so much variation in results from different testing companies?

[Angela]

See Razib Khan's take on the matter, which has credence particularly since he helped create one of them.


https://gnxp.nofe.me/2017/08/29/why-do-percentage-estimates-of-ancestry-vary-so-much/

 

[mwauthy]

The article makes sense. I guess what was left out of the article that is still a bit confusing is the variation and bias in what different companies determine to be representative of that region.
I can understand different reference samples from a region skewing the results. However, if there is scientific and historical theory behind these results shouldn't the various programmers from the various companies creating the frameworks for each region have similar historical and scientific understandings?
Is it bad science or bad history for one company to treat all of France as a homogenous ethnicity as opposed to another company splitting it into various regions based on the various historical influences and scientifically different snps?

 

[Angela]

The article makes sense. I guess what was left out of the article that is still a bit confusing is the variation and bias in what different companies determine to be representative of that region.
I can understand different reference samples from a region skewing the results. However, if there is scientific and historical theory behind these results shouldn't the various programmers from the various companies creating the frameworks for each region have similar historical and scientific understandings?
Is it bad science or bad history for one company to treat all of France as a homogenous ethnicity as opposed to another company splitting it into various regions based on the various historical influences and scientifically different snps?

I guess my cynical side is telling me that if they use the exact same reference samples and algorithms then they don't have anything different to sell.

I think in terms of Nat Gen, Jovialis has explained on the Italian genetics thread that the written explanation, at least the old one, does make a distinction between northern and southern France. It would be malpractice, imo, not to say that, because the southwestern France (I think it's Gascony) academic sample has already been analyzed and it's quite different.

A lot of the other things we've been discussing are judgment calls. You can't really say any of them are bad science.

I do think it's a huge mistake to use the Ashkenazim as a reference sample. They are anywhere from 40-60% "European", so it's going to screw up the analysis in my opinion. That was always my complaint about the FTDNA analysis and that of Dr. McDonald. The only way to really get a handle on whether or not you have Ashkenazi ancestry, imo, is to take the 23andme test. It will become extremely obvious based on the deluge of Ashkenazi "cousins" you get, and you'll even be able to figure out how far back it shows up. It doesn't quite go back to the 1200s with that algorithm, but it will go back pretty far.

 

[davef]

Why is it that some parameters have such high geographical coverage? I read about the Western European component once from ancestry Dna and if I recall correctly it covers france, I guess parts of north Italy, and Britain. How useful is that? If someone scores a certain percentage of it, how would they know if it's from southern France or England?

 

[mwauthy]

I guess my cynical side is telling me that if they use the exact same reference samples and algorithms then they don't have anything different to sell.

I think in terms of Nat Gen, Jovialis has explained on the Italian genetics thread that the written explanation, at least the old one, does make a distinction between northern and southern France. It would be malpractice, imo, not to say that, because the southwestern France (I think it's Gascony) academic sample has already been analyzed and it's quite different.

A lot of the other things we've been discussing are judgment calls. You can't really say any of them are bad science.

I do think it's a huge mistake to use the Ashkenazim as a reference sample. They are anywhere from 40-60% "European", so it's going to screw up the analysis in my opinion. That was always my complaint about the FTDNA analysis and that of Dr. McDonald. The only way to really get a handle on whether or not you have Ashkenazi ancestry, imo, is to take the 23andme test. It will become extremely obvious based on the deluge of Ashkenazi "cousins" you get, and you'll even be able to figure out how far back it shows up. It doesn't quite go back to the 1200s with that algorithm, but it will go back pretty far.

Ancestry DNA has me at 25% Iberian which for their reference samples is basically similar to Basque and 3% Italy/Greece.

Natgeo has me at 3% Southern Europe.

FTDNA has me at 0% Iberian and 10% Southeastern Europe.

These discrepancies appear to be more than just the differences caused by different reference samples for a region. They appear to be huge fundamental differences in theory about what the ethnicity of a region should look like. It's odd to me that professionals could have such different interpretations of scientific and historical data.

 

[mwauthy]

Why is it that some parameters have such high geographical coverage? I read about the Western European component once from ancestry Dna and if I recall correctly it covers france, I guess parts of north Italy, and Britain. How useful is that? If someone scores a certain percentage of it, how would they know if it's from southern France or England?

A country like France has influences from Basques, Celts, Greeks, Romans, Vikings, Franks and other Germanic tribes. So saying you are French or Western European is not that informative on an ethnicity estimate.

 

[Salento]

From Genographic:
".... This information is captured from points across your entire genome, so we’re able to see both parents’ information going back six generations or more. Your percentages reflect mostly recent influences, yet they may also show some ancient patterns in your DNA due to mixing of groups from different regions over hundreds, if not thousands, of years. If you have a very mixed background, the pattern can get complicated quickly. Use the reference population matches below to help understand your unique results."

 

[Jovialis]

For national geographic, I wonder why Italian/southern European, is the only European component that mentions a specific ethnicity. The others are just Northwestern, Northeastern, Southwestern, etc. It is rather peculiar.

I look forward to Dr. Vilar's comments.

 

[Jovialis]

I think in terms of Nat Gen, Jovialis has explained on the Italian genetics thread that the written explanation, at least the old one, does make a distinction between northern and southern France. It would be malpractice, imo, not to say that, because the southwestern France (I think it's Gascony) academic sample has already been analyzed and it's quite different.

It's for the new one.

I would imagine someone from south France would get Northwestern European, Southwestern European (Iberia-like), Italian/Southern European (Italian-like), and West Mediterranean (Corsican/Sardinia) admixtures.

Someone from Northern France would have different admixtures, probably a little more similar to Scandinavians.

 

[Jovialis]

This is Lydia Ramsey's results of Nat Geo vs Ancestry.

Ancestry just says she's 90% Scandinavian. Nat Geo has regional breakdowns that are found within Scandinavians.

 

[Jovialis]

I do think it's a huge mistake to use the Ashkenazim as a reference sample. They are anywhere from 40-60% "European", so it's going to screw up the analysis in my opinion. That was always my complaint about the FTDNA analysis and that of Dr. McDonald. The only way to really get a handle on whether or not you have Ashkenazi ancestry, imo, is to take the 23andme test. It will become extremely obvious based on the deluge of Ashkenazi "cousins" you get, and you'll even be able to figure out how far back it shows up. It doesn't quite go back to the 1200s with that algorithm, but it will go back pretty far.

Tried my best to get a good screen cap of the Jewish Diaspora component that Nat Geo uses.

taken from this video:

https://www.youtube.com/watch?v=fjPhFFmLb-4

 

[Promenade]

I've taken many different autosomal tests and even though I've complained about the discrepancies between them they seem to share at the least general trends regarding deep or general ancestry(For example how much ancestry you received from EEFs and other ancient groups, the amount of Northwest or east European ancestry you have.) Others look as if they are just trying to model you in anyway possible based off where you would place in a PCA, like Germans being represented as a mix of Scandinivians with Balkan and Turkish ancestry or Italians showing up as a mix of British isles and Mid East ancestry.

Some Autosomal tests id written off as rubbish are really great for specific ethnicities though. For example I'd highly recommend DNAland(it's free, I'm sure most of you here know about it) for Blacks and Latin Americans or anyone with African ancestry. 23andme and Ancestry DNA give really general feedback on African ancestry while DNAland has been able to pinpoint the African ancestry in Salvadorans to the Senegalese which matches with historical data of where the slaves in the region originated.

 

[Jovialis]

Some Autosomal tests id written off as rubbish are really great for specific ethnicities though. For example I'd highly recommend DNAland(it's free, I'm sure most of you here know about it) for Blacks and Latin Americans or anyone with African ancestry. 23andme and Ancestry DNA give really general feedback on African ancestry while DNAland has been able to pinpoint the African ancestry in Salvadorans to the Senegalese which matches with historical data of where the slaves in the region originated.

My friend who's Hispanic did the same test as me, but I thought his results were pretty much on target. Mostly Iberian, a little Italian, a critical mass of Native american, with marginal amounts of West African, and Jewish diaspora. Nevertheless, they don't get anymore specific than that; other than the descriptions. I guess it confirms some of his ancestors may have been conversos, that immigrated to Latin America.

 

[Jovialis]

Here's an interesting video comparing the results of four DNA companies; FTDNA, 23andme, My heritage, and AncestryDNA.

Furthermore, he give's a good explanation as to why there are differences in results. He also addresses the difficulties that come up with trying to accurately determine ethnicity for autosomal DNA results.

He touches upon three important points:

1. What is Ethnicity?


2. You only have DNA going back 6 generations; 120 ancestors. You don't receive DNA going back any further.


3. Issues with assigning reference populations; lack of DNA from the past.

Thus, I guess we can see that it isn't necessarily the quality of the test. It is just that the science and information is just not there yet.

@ Angela, I see what you mean in regards to the other thread.

 

[davef]

My friend who's Hispanic did the same test as me, but I thought his results were pretty much on target. Mostly Iberian, a little Italian, a critical mass of Native american, with marginal amounts of West African, and Jewish diaspora. Nevertheless, they don't get anymore specific than that; other than the descriptions. I guess it confirms some of his ancestors may have been conversos, that immigrated to Latin America.

I guess part of what determines the quality of the test is your own ethnicity. It seems that nat geno is accurate for south Italians such as yourself since it got you at 90 percent and was able to figure out your friend's makeup as well. DNA land can make ridiculous guesses for certain people, but based on a post in this thread, it does well with hispanics. Living DNA seems to cater well to British and Irish.

 

[Promenade]

My friend who's Hispanic did the same test as me, but I thought his results were pretty much on target. Mostly Iberian, a little Italian, a critical mass of Native american, with marginal amounts of West African, and Jewish diaspora. Nevertheless, they don't get anymore specific than that; other than the descriptions. I guess it confirms some of his ancestors may have been conversos, that immigrated to Latin America.

Yes, I agree, these tests are usually very helpful for Latin Americans because it's easier to distinguish their African, European and Amerindian ancestry which are more clearly defined than the differences between individual European ethnicities. DNAland is especially useful since it goes farther than the standard West vs Central African split, it even breaks up New World groups if only a little.

 

[LeBrok]

Yes, I agree, these tests are usually very helpful for Latin Americans because it's easier to distinguish their African, European and Amerindian ancestry which are more clearly defined than the differences between individual European ethnicities. DNAland is especially useful since it goes farther than the standard West vs Central African split, it even breaks up New World groups if only a little.

Very good point.

 

[Jovialis]

Yes, I agree, these tests are usually very helpful for Latin Americans because it's easier to distinguish their African, European and Amerindian ancestry which are more clearly defined than the differences between individual European ethnicities. DNAland is especially useful since it goes farther than the standard West vs Central African split, it even breaks up New World groups if only a little.

http://www.eupedia.com/forum/threads/32730-Strange-Geno-2-0-results?p=508191&viewfull=1#post508191

This poster had some very interesting results with the same test. It was able to give him a nuanced breakdown. He was from European background.

 

[Promenade]

http://www.eupedia.com/forum/threads/32730-Strange-Geno-2-0-results?p=508191&viewfull=1#post508191

This poster had some very interesting results with the same test. It was able to give him a nuanced breakdown. He was from European background.

It is interesting, although I don't believe his 2 percent East African is from African Americans. Similarly as you noted that the East African is not accompanied by middle east nor North African ancestry, it is not accompanied by the central or west African ancestry you'd expect to find in someone with African American ancestry. The majority of Africans were brought over from West and Central Africa and it's doubtful that the religious, abolitionist minded Quakers mingled with them like slave owners did. My guess is it's related to his mtdna.

 

[Jovialis]

It is interesting, although I don't believe his 2 percent East African is from African Americans. Similarly as you noted that the East African is not accompanied by middle east nor North African ancestry, it is not accompanied by the central or west African ancestry you'd expect to find in someone with African American ancestry. The majority of Africans were brought over from West and Central Africa and it's doubtful that the religious, abolitionist minded Quakers mingled with them like slave owners did. My guess is it's related to his mtdna.

You have a good point about the west Africans being the source of African slavery in the United States. But I didn't know mtdna could contribute up to 2% for autosomal DNA. So perhaps that may not actually come from his Quaker heritage.

Nevertheless, there was still a time in colonial American history prior to miscegenation laws where Africans would intermix with non-Anglo European indentured servants (many of whom were Irish). You should check out the video I posted, he talks shortly about possible admixture with old-stock Americans. Which is also why I suspect there is marginal Native American admixture as well.

Miscegenation laws, forbidding marriage between races, were prevalent in the South and the West. Because English masters had had little regard for indentured servants of non-Anglo ethnic groups, they allowed and sometimes encouraged commingling of their servants. Being seen in public or bringing legitimacy to these relations, however, was not lawful. This is evinced by a court decision from 1630, the first court decision in which a Negro woman and a white man figured prominently. Re Davis (1630) concerned sexual relations between them, the decision stating, “Hugh Davis to be soundly whipt . . . for abusing himself to the dishonor of God and shame of Christianity by defiling his body in lying with a Negro, which fault he is to actk. next sabbath day.”⁴⁰

https://memory.loc.gov/ammem/awhhtml/awlaw3/slavery.html