Thursday, August 22, 2013

My Autosomal DNA Quandary

I have had my autosomal DNA tested by AncestryDNA, by FamilyTreeDNA, and by 23andMe.  All three provided a different answer to my question - what populations did my DNA come from?

My perceived DNA origins, based on my genealogical research is approximately (from estimated origins of my 128 5th great-grandparents):

68%  British Isles
24%  Germany
1%  Holland
1%  France
6%  Unknown (perhaps some are French, some are English)
Here are my findings to date:

1)  AncestryDNA - I wrote about my results in Results from Ancestry.com Autosomal DNA Tests - Post 1, and Results from Ancestry.com Autosomal DNA Test - Post 2.  Here is the chart:



The chart says that I have 94% "British Isles" DNA and 6% "Uncertain."

The 6% "Uncertain" may include all of my perceived western European origins.  The 94% "British Isles" may include a significant portion of my western European origins - Angles, Saxons, etc.

AncestryDNA does not, yet, provide chromosome number and region details about the matches found with another AncestryDNA tester.  It does provide a comparison of surnames and a user can usually see the family tree of persons they match with.

2)  FamilyTreeDNA - I wrote about my FTDNA results in My Family Finder Autosomal Test Results - Post 1 and My Family Finder Autosomal Test Results - Post 2.  The chart is similar:


My guess is that the 10.78% "Middle East" component may include some of my Germany origins, and may indicate that some of them have Ashkenazi heritage.

Using the "Family Finder" results and matches to portions of data on each chromosome, I can see which chromosome I share with another FamilyTreeDNA tester.  That is very useful.  I can see the family tree of most FamilyTreeDNA testers to see where matches might occur.

3)  23AndMe - I wrote about my 23ndMe autosomal test results in My 23andMe DNA Test Results - Post 1, My 23andMe Test Results - Post 2: Chromosome Views UPDATED! and My 23andMe DNA Test Results - Post 3: Finding Matches and Seeing Connections.  The "Ancestry Compositon" chart looks like this:


23 andMe has a "Chromosome View" chart which shows the areas on each chromosome from each of the 22 populations identified worldwide:


This chart reflects what my genealogical perceptions are, with only a small exception - the 0.6% that is "East Asian and Native-American."  I've guessed that this may reflect 1 out of my 128 5th great-grandparents being Native-American.  I think it's probably from a grandparent of the French-Canadian wife of Abraham James Kemp, named Sarah Fletcher (a 3rd great-grandparent). 

Using the "Family Inheritance: Advanced" results provides visible matches to the chromosome portions I share with another 23and Me DNA tester.  That is very useful.  I can see the family tree of only a few of the 23andMe DNA testers to see where matches might occur.

4)  My quandary is the inconsistency between the three test results.  I understand that the three testing companies are using different reference populations, and that those reference populations don't separate out every country of origin, but they generally separate out the continents of origin.  Is this just the case of higher expectations than can reasonably be expected?

Another quandary is this:  Did AncestryDNA, FamilyTreeDNA and 23andMe test all of the same chromosomal regions?   If so, then there is no point in transferring 23andMe or AncestryDNA results into FamilyTreeDNA's database.  

The URL for this post is:  http://www.geneamusings.com/2013/08/my-autosomal-dna-quandary.html

Copyright (c) 2013, Randall J. Seaver

   

9 comments:

Eileen said...

Randy, what a great post and so relevant. I wrote on my DNA quandary today, too. Please see my post at http://www.oldbonesgenealogy.com/dna-and-me/.

Jacqi Stevens said...

Randy, thanks for your thorough posts on the various facets of DNA testing you are observing. Though not quite so facile as you in handling the resulting data, I'm going through the genetic testing process on behalf of two relatives, myself--and sometimes feeling as if I need to return to college to get an advanced degree on the subject before proceeding with any analysis...or...confess that I feel mildly as if I am a player in a variation of the unfolding tale, "The Emperor's New Clothes." Are these results valid? Why the differences? You are the same "you" in all three iterations, I'm presuming. Like you, I have a lot of questions (even though I'm doing my best to get myself educated on the subject), but my questions are much more rough around the edges.

Anonymous said...


Randy, I don't know much about dna yet. I have found many things in 23andme.com 's testing to be correct in the relatives and health categories. I used their testing and watched it with a cynical eye. I am satisfied.

I am wondering who's right between 23andme and gedmatch.com or both.

Ed Thanks for your blog!

Cary Bright said...

Randy,
Instead of transferring your results, go to each site and down load the "raw data" from each autosomal testing company and then go to http://www.gedmatch.com to upload each of the raw data files with the wonderful upload directions you find after you create a Gedmatch.com login. You will have to wait a few weeks to see the comparisons, but soon you can see how you relate to your other "twins". Your fun is just beginning once you learn your Chr. matching!!

Nathan Machula said...

The results would only be inconsistent if they were using the same reference populations and methodology, which they're not. None of them adequately explain what the results actually mean for a lay audience, but some do a better job of it. Ancestry's methodology is still a black box. At the very least, the labels they've assigned to the genetic components have serious problems.

FTDNA's methodology is mostly explained in their FAQ, although being spread out as answers to various questions makes it somewhat difficult to understand in a coherent manner. It does answer the question you raised about using "the same chromosomal regions" from their perspective: they only use 295k SNPs out of the 710k tested. Many of the SNPs either don't differ much between populations, i.e. not ancestry informative markers (AIMs), or they're typically aligned with other SNPs and thus redundant, i.e. linkage disequilibrium. The underlying algorithm they use was designed by Dr Doug McDonald, who's explained some of its quirks in various forums.

23andMe has been the most transparent about their methodology. They have a rather detailed page on their website, plus an even more technical poster they displayed at the annual ASHG meeting.

All three autosomal DNA tests use the same base chip. 23andMe just added a few hundred thousand SNPs beyond the ones in common. The data from each should essentially be identical, though each may use different SNPs in their bio-geographical ancestry (BGA) analysis. If you've already had the FF, there's no reason to import any autosomal data at FTDNA.

Overall, no BGA analysis is particularly useful for genealogy. Instead, figuring out how genetic matches are related can help support or refute traditional genealogy research.

Anonymous said...

Hi Randy,
I second Nathan's opinion. I myself have tested at Ancestry, 23andMe, and Nat'l Geographic's Geno2. NatGeo gives pretty good info as to their comparison populations; for example, they compared me to Danish and Tuscan populations (and my mom has Danish and Italian roots).

It really is all in the comparions population. You might like checking out dodecad.blogspot.com; the author provides various "admixture" populations which you can use on your own PC to compare your downloaded DNA results.

Each "admixture" group he provides which I've tested against gives me different results, because the test populations vary.

Still, though, I generally end up as European, for the most part. Which is good, because that's my heritage! :)

Best regards, Cathy D

Anonymous said...

Read Ancestry's white paper (don't worry abt the math parts)if you want their "process" on ethnicity estimates and realize that none of these companies are digging up ancient humans to get "Clean" DNA. Emigration, immigration, invasion and prehistoric/historic population movements make finding "clean" DNA REALLY hard. Also keep in mind that 23andMe look like the test more markers that you might assume help you with genealogy when in fact they are looking at genes associated with medical issues. Test with more than one company (if you can afford it) and accept the fact that science moves faster every year so if you repeat your testing every, say, three years, you may "look" quite different from an ethnicity perspective.

Kitty Cooper said...

GEDmatch has many different ancestry calculators plus if you have a parent tested you can phase your results there and look at the phased ancestry. I did a post about GEDmatch here http://blog.kittycooper.com/2012/12/what-to-do-at-the-gedmatch-site/ that you might find useful

Anonymous said...

randy, I second what cary bright suggested. there is another service that performs SNP marker analysis though. it is called DNA Tribes, and the SNP analysis costs only $40 or $50. you just download the raw data from the services that you received your results through, then you e-mail it to DNA Tribes and it takes about a week or two for them to analyze it. also, it would probably be a good idea to study the results from ancestryDNA and 23andme to find out the percentages of the different haplogroups that you inherited, then look for maps that show the regions of Europe where those haplogroups are most common, because some of them are common in more than one European country.