Wednesday, March 6, 2013

Who Do They Think I Am - A Look At Four Autosomal Analyses


Since a number of genetic genealogists have already participated in the exercise of analyzing their results from the various autosomal companies, I have decided to look at mine as well. To see what others have discovered, see the posts by CeCe Moore and Roberta Estes. In my analysis, I will only look at the results from four commercial entities that provide autosomal results: National Geographic’s Geno 2.0 Project, Family Tree DNA’s Population Finder, Ancestry.com’s AncestryDNA, and 23andMe’s Ancestry Composition.

Each of these four companies provided different results and I will compare these in light of what I know concerning my own ancestry from the last 500 years. Attempting to assign a person to a population is less difficult for someone who has a homogenous ancestry than it is for someone who is admixed from divergent populations. Some of the services will assign a primary population while others look at the constituent parts of one’s genetic background and provide an analysis of the segments.

Regional Populations


Another problem in comparing the results is that the various companies use different reference populations. In addition, regional populations are not consistent. For Europe (where all of my known ancestors hail), 23andMe classifies four regional populations: Northern European, Southern European, Eastern European, and Ashkenazi.

FTDNA’s European regions are identified as Western European, Northeast European, Southeast European, and Southern European. AncestryDNA features more European regions and these include British Isles, Scandinavian, Central European, Eastern European, and Southern European.

The Geno 2.0 project assigns local results based on a mixture of a variety of world regional populations with only two that are predominantly European in origin: Mediterranean and Northern European; however, the Mediterranean segment classification is not limited to Europe. With only two regional populations assigned to Europe, it is difficult to compare the Geno 2.0 results with the others – but we will get to this later.

My Ancestry


To the best of my knowledge, the following chart illustrates the nature of my ancestry within the last several hundred years. While I can take some lines back to the 1500s and beyond, others can only be traced satisfactorily to the early 19th century.

 
Primarily, I am English (38.28%) and German (31.25%). Scottish, Welsh, and Swiss are represented by each constituting 6.25% of my ancestry. My Scots-Irish, Irish, and French ancestries each contribute 3.13% of my lineage. My French ties come from the former province of Dauphiné in southeast region of the country.

Finally, my least represented known ancestry is of Norman stock from the Isle of Jersey. Two New England families on my father’s side constitute this lineage. My Gustin (formerly known as Jean de la Tocq) line and associated families are from St. Ouen’s parish and my Gavitt/Gavey line and related families hail from St. Saviour’s parish.

While I do not have any personal knowledge of Dutch ancestry, there are a number of residents of the Netherlands that match my mother on 23andMe with percentages that are consistent with third and fourth cousins. The origin of these connections has not yet been determined, but probably will show as one of my lines previously believed to be German. In addition, it is thought that my Maneval line, which originated in Dauphiné, may have intermarried with Italians in Piedmont.

National Geographic’s Geno 2.0 Project

In the Geno 2.0 project, the various reference populations are viewed from their specific admixture. Since my ancestry is European, we’ll concentrate on those references for this discussion. There are 12 reference populations from Europe and include the following ethnicities: British, Bulgarian, Danish, Finnish, German, Greek, Iberian, Romanian, Russian, Russian Tartars, Sardinian, and Tuscan.

All of the above populations have Northern European, Mediterranean, and Southwest Asian components. Certain populations (Bulgarian, Finnish, Romanian, Russian, and Russian Tartars) also carry segments from Northeast Asia. Depending upon the reference population’s geographic location, the majority of the segments were either Northern European or Mediterranean.

Mediterranean is also found as the majority component in the following non-European populations: Egyptian, Georgian, Iranian, Kuwaiti, Lebanese, Northern Caucasian, Puerto Rican, and Tunisian. Mexican-Americans also have a sizable Mediterranean component; however, Native American is their greatest percentage.

Other regional reference markers that are not found in the European reference populations are Southeast Asian, Native American, Oceanian, Subsaharan African, and South African. For an overview of the reference populations used in this study, go to https://genographic.nationalgeographic.com/overview-of-regions-and-closest-populations/reference-populations/.

For my results, my Northern European component at 41% is less than the Northern European reference populations of Finnish (57%), Danish (53%), Russian (51%), British (50%), and German (46%). My Mediterranean component (39%) is greater than that which is found among German (36%), British (33%), Danish (30%), Russian (25%), Russian Tartar (21%), and Finnish (17%). It is also considerably less than more southerly European populations from Sardinia (67%), Tuscany (54%), Greece (54%), Iberia (48%), Bulgaria (47%), and Romania (43%).

Since European populations also have Southwest Asian genetic components, my 19% is slightly higher than most of Geno 2.0’s European reference populations; however, it appears to be more closely aligned with Eastern Europeans such as Russians (18%), Romanians (19%), Bulgarians (20%), and Russian Tartars (21%); however, I do not have any Northeast Asian markers, which are characteristic of all of these populations.

I have included a chart of four reference populations compared to my results. Included in those four are the primary (German) and secondary (Tuscan) reference populations as determined by Geno 2.0. I have added two additional populations (British and Danish) for comparison purposes.


Geno 2.0 lists German as my primary reference population. I am in agreement with this as I have a large percentage of German ancestry and an even larger percentage of English. When one remembers that Saxon, Angle, Jute, Frisian, Viking, and Norman invasions occurred on British soil, Germanic segments would have contributed greatly to this ancestry.

According to Geno 2.0, “This reference population is based on samples collected from people native to Germany. The dominant 46% Northern European component likely reflects the earliest settlers in Europe, hunter-gatherers who arrived there more than 35,000 years ago. The 36% Mediterranean and 17% Southwest Asian percentages probably arrived later, with the spread of agriculture from the Fertile Crescent in the Middle East over the past 10,000 years. As these early farmers moved into Europe, they spread their genetic patterns as well. Today, northern and central European populations retain links to both the earliest Europeans and these later migrants from the Middle East.”

Geno 2.0’s secondary population for me is Tuscan. Even eyeballing the results tells me something is amiss. While I have a Mediterranean percentage that is larger than the Northern European references, it is not comparable with those from Tuscany. I have included British and Danish references in the above graphic and they appear to be more in line with secondary and tertiary populations.

If I were to score the populations based on the total percentage differences of the three categories of Northern European, Mediterranean, and Southwest Asian, the Tuscan reference is not as close as the British and Danish references. I have a total point difference of 10 with the German reference; however, the Tuscan population has a 30 point spread.

British, which is logical from what I know of my own ancestry, only has 16 points of difference, while Danish is further removed with 24 points of difference – still less than the Tuscan example. While I would be in agreement with the Germanic identity, I am not in agreement with the comparison to Tuscan populations.

Family Tree DNA’s Population Finder


When I first received my Population Finder results, I immediately dismissed these because of the inclusion of 8.42% Middle Eastern ancestry. My Western European ancestry was reckoned as being 91.58%. Knowing that my lineages were all European, I could not see where Middle Eastern segments could exist within the past 500 years; any Middle Eastern ancestry would certainly been too far removed to show in my analysis. Since receiving the Geno 2.0 results and seeing how pervasive Mediterranean and Southwest Asian segments were across all European populations, I have rethought my original opinions on these results.

Since populations are more complex than I originally thought, I am more inclined to view the Middle Eastern segments as either part of what is identified by Geno 2.0 as either Mediterranean or Southwest Asian in origin. This remains to be seen and since neither service provides a chromosome by chromosome analysis, it is impossible to see if there is a correlation.

Ancestry.com’s AncestryDNA


Ancestry’s analysis has me baffled, as they have assigned 21% of my ancestry to Eastern Europe. While I have a slight amount of my overall lineage traced back to Ukraine, it was 38 generations in the past and its overall impact on my autosomal results should be negligible.

 
Although not all of my ancestors are represented by the pins on the map shown below, the ones that are present show the predominance of my heritage coming from the British Isles and Central Europe. None are found in Eastern Europe. While I would love to lay claim on some recent Slavic ancestry, I cannot and I question the results as reported by Ancestry.com. Like Geno 2.0 and FTDNA’s Population Finder, Ancestry does not plot the results by chromosome.


23andMe’s Ancestry Composition


Introduced in December 2012, I will have to admit that 23andMe’s new feature is far and above the competition in accuracy based on my known ancestry. In the previous incarnation called Ancestry Painting, 23andMe’s ancestral analysis was pretty Spartan. My results were, in a few words, pretty vanilla – or in the color schemes used at the time – completely blue.


The new Ancestry Composition feature fine tunes these results with additional global populations going beyond their original European, Asian, and African classifications to the expanded European, East Asian/Native American, Middle Eastern/North African, South Asian, Sub-Saharan African, and Oceanian regional populations. In addition, several sub-regional populations were also added.

 
23andMe also defaults to a standard estimate of your populations and allows you to determine if you want to be more speculative or more conservative in your population estimates. I’m ready to go for broke (read “reckless abandon” for me) and completely rely upon the speculative results as it gives me more options.



While these results remained 99.7% European, some additional colors were added to my ancestral spectrum. These were very small by comparison with 0.1% each for Native American, North African, and South Asian. The results could be regarded as noise or just very small segments of my ancestry.

With two of these populations occurring at the same segment as my mother, I have a tendency to believe that they may be accurate – but very persistent and fairly distant markers. She shares the Native American and North African segments. Therefore, the Southeast Asian must come from my father. How they fit into my ancestry, I haven’t a clue. I would have thought that my father’s side had more of an opportunity to have Native American blood, as a majority of his ancestors were in the colonies over a hundred years prior to my mother’s first immigrants.



One thing that I believe is incorrect is the assignment for my X chromosome as being “British and Irish.” Having already phased my X as coming primarily from my maternal grandmother (see my previous post on this subject), I already know that her ancestry was 87.5% German and 12.5% French; however, the contributors to her X chromosome were all German.

Outside of this misidentification, I am pleased with how 23andMe assigned the various populations. My German ancestry is somewhat underreported; however, I am assuming that most of what came from my Teutonic predecessors is found under the “Nonspecific Northern European” category. While not having any known Sardinian (0.3%) or Balkan (0.2%) ancestry, I checked with my mother’s results and found that she only shared the Balkan markers. She also had a chromosome that was nearly all Italian which I did not inherit. This may indicate the supposed Italian ancestry from Piedmont.

The Sardinian must come from my father; however, his ancestry was primarily from the British Isles. There is one possibility though. My grandfather’s sister’s middle name was the Italian surname of Marcelli. Unfortunately, we have no clue why the second child of this family was named Essa Marcelli Owston. Was she named for a Sardinian or Italian ancestor or a friend of the family? Of my great grandparents’ five children, this is the only name that cannot be traced to a family member or a friend of the family. Alas, this is another mystery that hopefully can be solved at some time in the future.

Conclusion


Of the four autosomal services, I would have to say that 23andMe has the best analysis and it lines up closely with my known ancestry. It is the only service that drills down to the sub-regional populations and gives you the opportunity to speculate or be conservative about the analysis. It is also the only service that provides a chromosomal analysis. With the current price at $99, if you are looking for an inexpensive ancestral analysis, 23andMe is the route to consider.

16 comments:

  1. Nice analysis Jim! Thank you for the mention too.

    ReplyDelete
    Replies
    1. Thanks CeCe - you did this before anyone else and deserved the mention.

      Delete
  2. Excellent job Jim! Going to follow you ( I too am at Geneabloggers) and share your Blog on my other sites.
    Katie

    ReplyDelete
  3. So it is true that not every percentage someone gets is definite. So the "1% or 2% Native American/East Asian" percentage that every African American gets may not be real after all?

    ReplyDelete
    Replies
    1. Smaller segments are usually suspect, but at 1% and 2%, these probably real. Less than 1%, then it is in the realm of probability.

      Delete
  4. Very comprehensive analysis. I've had my autosomal DNA done by three companies and the results disagree widely. Ancestry has me as 63% Scandinavian; 23andMe says 0.99% Scandinavian (Finnish); and Family Tree DNA assigns 12.23% to Tuscan/Finnish/ Romanian/Sardinian but does not break it down. I wish the companies could agree on a standardized reporting system. I contacted Family Tree DNA about the differences. Their answer is that the results do not really vary that much: the companies use different standards. I wish they would agree on one standard so we would be better able to understand and use the information.

    ReplyDelete
    Replies
    1. Jackie:

      Amen to that. It is almost like Beta and VHS.

      23andMe should show two Scandinavian categories - one for Scandinavian (i.e., Norse) and one for Finnish.

      Jim

      Delete
    2. This comment has been removed by the author.

      Delete
    3. Your comment confuses me in trying to make a decision which company my wife should use. She is overwhelmingly ethnic Finnish (i.e. not from the Swedish speaking Finnish origins) but has one line of ancestors originating from Sweden seven or eight generations back. It is thus important to have a clear differentiation between them. You state 23andMe should show separate Nordic and Finnish categories, but Jim's example shows that is does. Can you clarify what you mean?

      Delete
  5. This is one side of the story;
    - 23's Ancestry Composition can not distinguish between Northern Euopean and Eastern European. People from Germany get 100% Northern European, while their East-European matches get 100% East-European.........
    - For many people Split View does not work.
    - far too little DNA samples from other continents

    ReplyDelete
    Replies
    1. If the markers for Northern European and Eastern European are different, I'm not sure what the problems are.

      Correct me if I am wrong, but I think in split view you have to have both parents tested.

      Delete
  6. I manage an account of someone from Germany who has many East-European matches. He scores 100% Northern European, but most of his East-European dna matches also score 100% Eastern European. So then where does their common ancestor came from ?

    For many people WITH BOTH PARENTS tested, Split View does not work.

    Ancestry Composition is not working properly, especially not for Europeans. 23andme do not care, because their business strategy is focused on the health care market. In time customers will find out that Ancestry Composition does not work properly.

    ReplyDelete
    Replies
    1. Thanks for the further explanation. Unfortunately, I don't have an answer for you. I wasn't aware that split view was not working for individuals with both parents having tested. Unfortunately, I only have one parent tested and have not tried it.

      Delete
  7. I'm just getting ready to purchase an autosomal DNA kit and have been researching everything so I can decide. Do you think your review would be current for today? I don't know if any of the companies have changed their testing or analysis since this blog was written.

    ReplyDelete
    Replies
    1. Debbie: All three have updated their results; however, my estimation still rates 23andMe more like my actual relationship.

      Delete