Tuesday, December 18, 2012

Comparing Admixture Test Results Across Companies (otherwise known as "ethnic" breakdowns): FTDNA, AncestryDNA, 23andMe and Geno 2.0 - My Review

With the recent releases of Geno 2.0 and the new 23andMe Ancestry Composition feature, I thought it would be a good time to review and compare my biogeographical ancestry (BGA) results from each of the major DNA testing companies. (BGA is sometimes referred to as ethnic breakdown or admixture analysis.)

My Known Ancestry
First, let's take a look at what I know of my ancestry:
25% Finnish
12.5% Scandinavian (Norway)
12.5% British (England)
~25% Colonial New England (presumably mostly British)
~15.63% Unknown (due to brick walls and immigrant ancestors of unknown origins)
~9.38% German

CeCe's Ancestry Percentages
You can create your admixture pie chart here.

Family Tree DNA's Population Finder
FTDNA's Population Finder (comes with Family Finder) was one of the first commercially available admixture tests based on chip technology.

I am listed as 98.93% Western European (Orcadian) with a 1.31% margin of error. Surprisingly, none of my substantial Finnish ancestry was detected. In my case, Population Finder is not very informative so I am glad that FTDNA will soon be updating this feature. Of course, many people's Population Finder results will be much more detailed than mine, so my experience may not be a typical representation of this feature's capabilities.

AncestryDNA's Genetic Ethnicity
Just a couple of weeks ago, AncestryDNA's admixture tool was the newest entry into the market, but times are changing fast.

It's one thing for results to not be especially informative like Population Finder above, but for them to be downright wrong is disappointing. This feature underestimates my Finnish ancestry and overestimates my Scandinavian ancestry, both substantially.  Because I have tested my mother, I know that they are estimating a large amount of Scandinavian from my mother's side when there is no Scandinavian ancestry in her extensively documented family tree at all. Does that mean that she doesn't have any? Not necessarily, but AncestryDNA estimates that she has 62% Scandinavian and only 23% Finnish (instead of ~50%) ancestry. Although none of these admixture tools are perfect, this one seems to be the most misleading. Notably, I have received many messages from people confounded and even very upset by their AncestryDNA Genetic Ethnicity results.

23andMe's Ancestry Composition
23andMe started out with a very conservative three population breakdown called Ancestry Painting. Customers had long been requesting an updated version. Happily, the brand new Ancestry Composition is much improved. There are three different views and a number of resolution and confidence settings. I prefer the most granular (Sub-regional Resolution) and lowest confidence (Speculative) of those.

Map View:

Chromosome View:

Split View:

This version matches my known ancestry the most closely and is also in line with the ancestral origins represented in my Relative Finder/Family Finder/AncestryDNA match lists. It was fun to see that even my tiny amount of Ashkenazi Jewish ancestry was detected. Although this tool is definitely the most informative and flexible, it isn't without problems in this early version. For example, on the Split View, all of my Finnish should be from my maternal side, but you can see that 4.6% of my estimated 21.5% Finnish is erroneously assigned to my paternal side. Conversely, almost all of my Scandinavian is correctly assigned to my paternal side. Since my deceased father wasn't tested at 23andMe, they had to compute the Split View without my father's results, which makes it more challenging. However, since I have tested both of his siblings there (they don't have any Finnish), advancements in the future may help in this regard.  I have many relatives tested at 23andMe and the vast majority of them have results very much in line with their family trees. Both of my sisters' Ancestry Composition results seem even more accurate than mine, as do my mother's, both of my aunts' and my uncle's. (I may just be a bit of a mutant!)

National Geographic Geno 2.0's Who Am I?
As most of you know, the new phase of the Genographic Project started returning results a couple of weeks ago.  Screen shots show the different components that make up my ancestry according to these results.

Then, they list the two reference populations that match the combination of my components the most closely. Interestingly, even though I am only one quarter Finnish, they chose Finnish as the first one.

Next, they choose Iberian (Spain and Portugal) as the second population, which doesn't fit with any of my known ancestry.

This admixture tool is different than the other ones reviewed here because it is looking at deep ancestry from thousands of years ago. Personally, as a genealogist, I don't find it as informative or as satisfying as the admixture results that reflect more recent ancestry. I also don't really like how each person's unique combination of ancient components is essentially forced into a single (more recent) population label. This, no doubt, works very well for those whose ancestors have been in one geographic region for hundreds or even thousands of years, but it does not result in highly relevant results for those of us who are highly admixed with ancestry from many different regions. I find the tools that analyze and attempt to categorize small portions of my DNA separately, giving percentages of each region or ethnic group that add up to my total ancestry more meaningful.

At this point in time, for those interested in the origins of their ancestors in the last couple hundred to few hundred years, 23andMe is the best choice in my opinion (especially at the new price of only $99). If your interests are more focused on deep ancestry and anthropology, then Geno 2.0 would be my recommendation. (Remember though that this review is focused on admixture features only. There is much more to take into consideration when purchasing a DNA test for genealogical purposes!)

All of these companies would be well-advised to focus on improving their admixture results for those with non-European ancestry. (Obviously, the majority of their customers have been of primarily European ancestry so far, but optimally everyone will have the the same opportunities of self-discovery in the future.) 23andMe has publicly stated their plans to do so and I'm sure that the other companies will address these shortcomings as well.  AncestryDNA's acquisition of the Sorenson database and National Geographic's vast collection of indigenous population samples puts them in the best position to advance in these areas in the near future, but as 23andMe has demonstrated, customer samples carefully vetted with genealogical information can be utilized with great success as well. As more people are tested, particularly internationally, there will be improvement across the board.

It is important to remember that as intriguing as these admixture predictions are, none of them are 100% accurate at the granular level. We still have a long way to go before anyone can honestly claim to be able to tell a person exactly where their ancestors once lived based on their autosomal DNA alone. However, we are making progress and in a couple of years we will likely be amazed at the advances. In the meantime, the ever-increasing competition between the major companies is proving to be beneficial, spurring all to improve their offerings. So, don't count any of them out quite yet. In a short time, my opinions regarding the "best of" may have completely changed again!

*For more information on these features, please see:

[Disclosure: I have relationships with all of the companies reviewed above. I received complimentary Geno 2.0 and AncestryDNA kits for review purposes. I occasionally receive complimentary 23andMe test kits through my work as a volunteer "23andMe Ancestry Ambassador". Additionally, I have affiliate relationships with Family Tree DNA, 23andMe and AncestryDNA and own a small amount of Ancestry.com stock. I always strive for complete objectivity - none of these relationships have affected the opinions stated in this blog in any way.]

Wednesday, December 12, 2012

My Geno 2.0 Results: Step-by-Step

As much fun as I have had posting and reading about other people's Geno 2.0 results in the last couple of weeks, I have to admit, there's nothing quite like getting my own (finally)!

According to the "infographic" below, I am one of 559,515 Genographic Project participants. (You can access this individualized image through the yellow "Share" button on the top right of the "Your Story" page.)

There were 524,384 participants from Geno 1.0, so judging from this, there are now 35,131 new Geno 2.0 participants. That is certainly a lot and we don't even know if that is the number of kits sold to date or just the number of results being returned now. This means at least 35,131 kits have sold since Geno 2.0's introduction in July of this year. (Pretty awesome!)


My Geno 2.0 mitochondrial DNA haplogroup is U5b1b2 which is consistent with 23andMe's and mtDNA Community's label for me, while my full mitochondrial sequence at Family Tree DNA designates me as mtDNA Haplogroup U5b1.

If you click on the arrows on your map, you will be walked step by step through the migration pattern of your direct maternal line ancestors, ending with a heatmap of the frequency of your subclade (if available). Of course, being female, I don't have a Y-chromosome to explore, but as part of their results, males also receive their Y-DNA haplogroup subclade (terminal SNP), direct paternal line's migration route and heatmap.

The next step that is recommended by National Geographic is to "Complete Your Profile" and "Contribute Your Story". These can be accessed through the Profile tab and the Our Story tab respectively, but for the first to be accessible, you need to opt into research participation under Profile > Account Settings. The default is "You are currently not participating", but if you check the box below and click on save...

...it changes to this:

Then, under Profile, go to "About Me", "About My Family" and "About My Ethnicity" and fill in the pertinent details.

Next, go to the Our Story tab at the top and you will see this:

If you scroll down below this, you will find this window, where you should enter information about your direct maternal ancestral line. This story should only include information on your mother's mother's mother's (etc...) line. Here's mine:

To see your story and the others that have been contributed, click on the "Read Stories" button on the opposite window under "Browse All Stories".

These are the other participants whose mtDNA haplogroup is U5b and have contributed their stories so far. There aren't very many people tested with mtDNA like mine yet as you can see from the "Universe" graphic above (the big blue circle with the red-orange center). The five dots are people whose mtDNA is most similar to mine.

Just for fun, here is peek at a few of the public Y-DNA stories. See anything interesting?

Next, let's take a look at my autosomal admixture results.

According to this, my admixture includes:
45% Northern European
35% Mediterranean
15% Southwest Asian
2% Northeast Asian

Which places me closest to...

Pretty cool since I am 25% Finnish, which as far as I know, is my biggest chunk of ancestry from any single area. My percentages don't match up exactly by any means. My Mediterranean component is significantly higher and my Northeast Asian component is lower than the typical Finn. The description for this latter component notes "... it is also found at a frequency of 5-10% in the Finns, likely introduced by the migrations of the Saami people from Siberia into Finland over the past 5,000 years." Since only one quarter of my ancestry comes from Finland, this discrepancy makes sense.

But, wait, hold on...

Iberian?! I don't have any known Iberian ancestry. Anyway, it doesn't look to me like I match it all that closely anyway.  (For details on how they reach these conclusions, read my earlier post.)

I'm not sure that this method of trying to match all of a person's ancestry to one population label works well for very admixed individuals like me. My individual components may appear to fit best with these two populations when taken as a whole, but this doesn't account for the mixed ancestry I have from multiple regions.

This part of the test is definitely intriguing although I don't really know what to make of it.

The Neanderthal percentage is very close to my 23andMe Neanderthal result of 2.5%. The Denisovan seems on the higher end compared to other results I have seen, but investigating that will have to wait for another day.

There seems to be much confusion regarding downloading the raw data file and transferring the results to Family Tree DNA, so I thought I would review the Expert Options section.

To transfer your results to Family Tree DNA, go to the Profile tab and then Expert Options:

Click on "Transfer to FTDNA":

Check the option for Geno 2.0 and enter your NatGeo Kit Number (found on your box and/or Profile> Account Settings> Geno 2.0 ID Code, where you have previously entered it to register). Then, check the box if you have a Family Tree DNA kit and enter your kit number and password as above. Click on "Next". On the next screen, you will be prompted to enter your address and it will look like you are going to be charged, but choose "Invoice" (instead of Credit Card) and keep going. Then, you will receive a screen to review that will show a cost of $0. Place the order and, if successful, you should get this:

and this:

 If you aren't sure if it worked, check your home page for this:

So far my account results don't show anything different, but I already have the mtDNA full sequence, so I'm not sure what would be imported anyway.  Many of the men importing their results are getting an extensive list of Y-SNPs on their Haplotree page like this:

Although importing Geno 2.0 results doesn't delete the results from the SNP testing that was previously performed at FTDNA, it will override what appears on the project pages.

Apparently, there is a delay for some accounts receiving the raw data download option and so far, mine hasn't appeared yet. When it does, it will be located under the "Expert Options" tab, just above the "Transfer Data to Family Tree DNA" option and look like this:

It is downloaded into a CSV file after clicking on the Download button.

I was hoping to be able to check my raw data file to see how my mtDNA heteroplasmy was reported, but apparently that will have to wait for another time. 

Obviously, I don't have my own Y-DNA results to review, but I have been reading all of the reports from our citizen scientists who are already immersed in investigating those newly released results. I will be sure to report on their findings since this will, undoubtedly, be the area of the most groundbreaking discoveries.

In my next post, I will compare and review my admixture results from all four companies.

[Disclosure: I received a complimentary Geno 2.0 kit from National Geographic for review purposes. This has not affected my opinions or analysis.]

Tuesday, December 11, 2012

23andMe Receives $50 million and Drops Price to $99 in Bid for One Million Customers

23andMe Raises More Than $50 Million in New Financing
The good news just keeps coming in the genetic genealogy world!  Just think what this will mean for all of our research...a pool of ONE MILLION customers to fish for relatives! If you have been wondering what to get for the holidays, here it is!

Company Sets Growth Goal Of One Million Customers, Reduces Price to $99 from $299

MOUNTAIN VIEW, Calif. – December 11, 2012 – 23andMe. Inc., the leading personal genetics company, today announced it has raised more than $50 million in a Series D financing. Participants in the financing include Yuri Milner, a new investor, as well as existing investors Sergey Brin, 23andMe CEO Anne Wojcicki, New Enterprise Associates, Google Ventures and MPM Capital. This investment will help the company achieve its growth goal of one million customers.

The Power of One Million People
Expanding the company’s ability to reach and serve one million individuals supports 23andMe’s goal to revolutionize health and wellness. It also will accelerate 23andMe’s ability to create a powerful platform that enables researchers around the globe to make meaningful discoveries significantly faster than is currently possible. With this expansion, 23andMe, which currently has more than 180,000 customers, will aim to:
  • Enable groundbreaking research by creating an exponentially larger collective of actively engaged, genotyped individuals;
  • Help accelerate development of new treatments;
  • Improve understanding of wellness and disease prevention; and
  • Broaden access for people seeking to manage their health and well-being through direct access and greater understanding of their own genetic data.
“A community of one million actively engaged individuals will be transformational for research. A community of this magnitude will improve researchers’ ability to quickly answer questions about genetic function and the role of environmental factors. In addition, it will enable researchers to understand medication efficacy and side effects, in both medications that exist today and medications are that are in development,” Wojcicki added.

Broadening Access: Lowering Price to $99
The Series D investment, combined with rapidly decreasing costs associated with genetic testing technologies, enables 23andMe to reduce the price of its Personal Genome Service to $99, effective immediately. The company will continue to evaluate optimal pricing strategies.
The investment also enables 23andMe to expand the necessary infrastructure to support growth in its research and operational capabilities, including product development, genetic research, software development, recruitment and marketing.

About 23andMe
23andMe, Inc. is the leading personal genetics company dedicated to helping individuals understand their own genetic information through DNA analysis technologies and web-based interactive tools. The company's Personal Genome Service® enables individuals to gain deeper insights into their ancestry and inherited traits. The vision for 23andMe is to personalize healthcare by making and supporting meaningful discoveries through genetic research. 23andMe, Inc., was founded in 2006, and the company is advised by a group of renowned experts in the fields of human genetics, bioinformatics and computer science. More information is available at 23andMe.

Be one of the million that will change the world - order your $99 23andMe kit now!

[**Update: The price of upgrades from v2 to v3 has also dropped to only $95.]

Thursday, December 6, 2012

23andMe's New Ancestry Painting - First Look!

23andMe's long awaited Ancestry Painting 2 (now called Ancestry Composition) is finally here. I am extremely pleased with my results and those of my family.

The results are presented as a ring chart with settings for different confidence levels and regional detail. The easiest way to figure out how it works is to just click around, trying the different options.

To inspect more closely, click on any of my images to enlarge.

This is my Standard Painting with no regional detail (still 100% Booooring!):

After clicking on the rings or on the + sign, I get this:

 Clicking some more, I get this:

Then, changing from Standard setting to Speculative, I get this:

I vastly prefer the "Speculative" setting since it seems to fit very well with what I know of my ancestral origins.

The pie chart below reflects my known ancestry. It was created from an Excel spreadsheet made by John Tierney and available on his blog.

Ancestral breakdown based on my family tree

Compare to my new AP2 percentages:

My Finnish and Scandinavian percentages seem a bit low, but in my sisters' paintings, they look about right:

 My mother's also appears to fit with what I know of her heritage:

My dad's siblings have an interesting surprise in that we finally got some different colors:

It appears that perhaps I do have some African ancestors after all (in a relatively recent time frame)!

My brother-in-law's ancestry becomes even clearer with the new Ancestry Painting. The small amount of Asian disappeared and was replaced with an increased African component. From what I've seen during beta testing, this appears to be the case with many of those with African American ancestry.

Compare to his old Ancestry Painting:

Another person who is half Indian got this interesting result:

  Compare that to his old Ancestry Painting:

Much improved, right?

 Look at this chart:

Beautiful, huh?


I am happy to report that, although we did not get a chance to beta test this portion of the new feature, the chromosome painting, including the X chromosome, is now live.

My painting:

With my Finnish ancestry highlighted:

This is what it looks like with one Finnish parent:

*Note that the paternal and maternal chromosome orientation is not always consistent in placement (top versus bottom) in this chart.
This man has one Indian parent:

Pretty clear, isn't it?

And this a beautiful painting:

Don't you agree?


23andMe uses their own customized version of a phasing engine called BEAGLE to separate out the genetic components inherited from each of our parents, even without actually testing them.  However, if you have a parent or child in the 23andMe database, then your phasing will be improved. This will lead to more accurate assignments as well as more detail contained in your ancestral breakdown (i.e. Italian versus Unspecified Southern European). Therefore, if you are fortunate enough to have tested one of these close family members, you need to let 23andMe know about this by connecting their profiles to your Family Tree.

If you have at least one of your parents in the 23andMe database (and connected to your tree), then you get a third view called "Split View". This view will automatically show you which parent contributed which portions of your ancestry. 

23andme is currently using 22 populations. They can all be seen on my breakdown here:

In order to get more information on the origin of the reference populations, you can click on their labels and it opens up to this:

For some populations, 23andMe has a lot to go on:

and for some, not so much (as expected):

Obviously, we would all like to see more detail and variety in the African reference populations. For now, this is all there is:

The new Ancestry Composition is expected to continue to change and improve as 23andMe builds their database. As you can see, this is especially important for African Americans. In the announcement on the forum, 23andMe geneticist and Product Manager Mike Macpherson acknowledges, "But there’s still more to do. One important example of this is that the system is designed to be able to update the reference populations. As it is now, Ancestry Composition can dive into European ancestry, but it offers limited resolution into East Asia and Africa. We’re working now on an update that will split up the Eastern Asian and Sub-Saharan African populations."

The 23andMe white paper explaining how this new feature works tells us just how far they have come, "We compiled a set of 7,868 people with known ancestry, from within 23andMe and from public sources. That’s over 15,000 chromosomes, since every individual contributes a chromosome from both their mother and their father. This a big jump over the 210 individuals that powered our original Ancestry Painting feature." The paper further tells us that the public reference datasets that make up the "Public" numbers that you see above, include the Human Genome Diversity Project, HapMap, and the 1000 Genomes Project. It goes on to explain, "Populations may be inherently difficult to distinguish because of historical mixing, or we might not have had enough data to tell them apart. As we obtain more data, populations will become easier to distinguish."

At least for me, this is a very good start! In fact, I think this just might be the best admixture tool I have tried so far. What do you think?