Friday, January 13, 2017

New 23andMe Ancestry Timeline Tool

I logged in to my 23andMe account today and learned of a new [BETA] tool has been added to our ancestry reports. Described as an Admixture Date Estimator and aptly titled, "Your Ancestry Timeline," the feature estimates when your ancestor with 100% of a singular ethnicity (ie West African) might have existed. Notably this is the second such feature that has been added to our Ancestry Reports; the first was the ability to download our Ancestry Composition "ethnicity" segment data (see blogs by Kitty Munson Cooper and Roberta Estes). At present it's unclear if 23andMe has rolled out "Your Ancestry Timeline" to all customers. I will tell you my opinion at the close of this short blog but let met introduce you to the feature first:

(1) After you log into into your 23andMe account, click on your Ancestry Reports and then Ancestry Composition. Next scroll down and (if you have it) you will see "Your Ancestry Timeline" located right under your "Ancestry Composition" results. At the top of the feature you will see a horizontal list of circled numbers, which represents a generation. Immediately below this are corresponding dates at 30-year intervals (30 years = 1 generation) and presumably driven by the birth date you listed on your 23andme account --- for me Generation 1 starts at 1940 and goes back to Generation 8(+) at the year 1700. Below this you will see colorful modules, each one representing an ethnicity component present in your Ancestry Composition results --- each module is of a different length based upon the time and generation range your proposed ancestor with 100% of a certain admixture (ie West African) might have actually existed.  It looks like this:

(2) Next click on any of the ethnicity modules, a pop-up box will open showing the probable time when your ancestor with 100% of a singular admixture lived. I clicked on the RED module "West African" as seen here:

The pop-up box shows: "You most likely had a parent, grandparent or great-grandparent who was 100% African. This person was likely born between 1880 and 1940." 

(3) To learn how the new feature works, you should (actually MUST) click on the blue link "Learn more about how to interpret this result" (located underneath the feature's ethnicity tabs) It includes a link to the 23andMe Admixture Date Estimator White Paper which goes into the "logic" and "science" behind the new feature.  The blue link reveals the following information: 

My Opinion

I like the 23andMe new Admixture Date Estimator ("Your Ancestry Timeline") feature because I use these sort  of statistical inheritance predictions when I analyze people's ethnicity admixture results. However my immediate impression is a lot of people will be confused and misinterpreting the information. To this point, the Admixture Date Estimator White Paper states the feature will "enable customers to find out, for each of the ancestries they carry, when they may have had an ancestor in their genealogy who was likely to be a non-admixed representative of that population." This is because genetic inheritance is much more complicated and ... well there has been very few ethnically "pure" or non-admixed populations in modern history. The population structure of most ethnic groups includes ancestry from "outside" groups. Also based on random genetic recombination we inherit any of our parents admixture in DNA chunks of varying sizes, with some of those DNA chunks remaining intact over several generations. This is problematic because the feature assumes a singular admixture component from one ancestor/ancestral event. Further note this estimator does not include your X-chromosome(s) in its calculations. 

Now I'm African-American, which means my admixture profile is going to be complex and colorful. I know that my West African admixture @ 78% comes from both of my parents and all of my grandparents in various amounts; this new feature automatically assumes my West African inheritance comes just one parent. In fact I descend from numerous indigenous populations living in western and other African subregions that entered my bloodline at different times and places in my genetic history. Further my results were phased with a parent and I've received "West African" from both of them in varying amounts. Therefore it is highly unlikely I had "a parent, grandparent or great-grandparent who was 100% West African that was born between 1880 and 1940." Or what if like me, you have Malagasy ancestry and you show a small Southeast Asian percentage @ 0.9%? Well Malagasy populations have been admixed with Asian and African for thousands of years so even if this Southeast Asian came one predicted ancestor he or she may not have been 100% Southeast Asian for a very long time. And what if you have Southeast Asian ancestry from both parents? 

Buried in the white paper (which most customers won't bother reading) states, "The admixture date provided is based on the ancestry segments estimated by Ancestry Composition, and is, consequently, dependent on their accuracy and specificity for accurate date estimation. Any genealogical history or ancestries that are not well captured by Ancestry Composition estimates may result in poor admixture date estimation, which typically results in older estimated dates of admixture. Secondly, the admixture date is based on all segments of a particular ancestry. If multiple genealogical ancestors contributed independently, the admixture date may reflect these multiple ancestors in a complex way. If many segments, from independent ancestors, recombine to form longer segments, the estimated admixture date may be shifted towards a more recent date. This is especially likely in the case when segments cover over 50% of a genome. On the other hand, if many older genealogical ancestors contribute discrete, shorter segments, the estimated admixture date may be pushed back, reflecting a weighted average over the multiple ancestors’ generations. Lastly, it is important to note that the inheritance of segments in one genome from a genealogical ancestor is a highly stochastic process, resulting in overlapping inheritance patterns that are not distinguishable the further back in time you go, even under otherwise ideal conditions. Thus, some amount of uncertainty is inherent in the data, so we present admixture date results in bins that allow for some of this inherent randomness." 

In my optimistic opinion these admixture date predictions may work better with people from more homogenous populations ("West Africa" is not one of them), or with smaller distinguished admixture amounts (ie Native American) that can be attributed to one ancestor (ie Choctaw grandmother) or ancestral couple.  Of course I'd much rather 23andMe work on improving/updating its Ancestry Composition (West Africa is a much too broad category) and Haplogroup predictions, including the build platform; restoring our ability to see the chromosome paintings of the people we're sharing with; implementing an in-house family tree utility, as well as expanding its fantastic In Common With/triangulation tool to include all of our DNA relatives. I would also like to see an ethnicity element added to 23andMe's chromosome browser/mapper tool where customers can see the location, size and ethnicity(s) of the shared segments displayed.  


No comments:

Post a Comment