View Full Version : 23andMe launches its new Ancestry Composition feature

08-12-12, 20:59
This much awaited new feature attempts to determine the percentage of total (autosomal) ancestry in 23andMe customers. We expected an detailed ancestry painting in the lines of an advanced Dodecad admixtures. There are in total 18 specific populations, among which 11 from Europe, and five additional nonspecific continent-wide groups. It's good, but it is considerably less than I expected. The most annoying is that some European members get a very high percentage of nonspecific Northern or Southern European, or an even more general nonspecific European.

I am also disappointed by:

the complete absence of Siberian and Central Asian categories
the lack of distinction within Balkan populations (Croatians share little in common with Albanians)
the lack of distinction between French and German (that is, Gallo-Roman vs Germanic, two root population that were originally completely distinct from each others)
the lack of distinction between Northeast and Southeast Asians (a Korean and a Thai both get 100% East Asian, which is pretty useless for them to know)
the lack of distinction between any group of Sub-Saharan African.

In these regards, the Dodecad admixtures are much more useful.

The Advanced Global Similarity tool on 23andMe had reference populations that are missing here, like the Basques or the Orcadians, which could have been interesting. I find it surprising that some reference populations are very specific (near isolates like the Sardinians, Finns and Ashkenazi), while others are aggregates of vastly different populations, like the Proven├žal French and the Saxons (under a common French-German category) or almost any Balkanese, Middle Eastern and African population.

In conclusion, it's an fairly interesting tool, but which has plenty of room for improvement.

08-12-12, 23:51
I don't know it but through your explanations, I find too it is very disappointing - almost a lost of time...

09-12-12, 02:46
I don't know it but through your explanations, I find too it is very disappointing - almost a lost of time...

the most accurate types of admixture tests for me is in gedmatch where JOlson has incorporated the databases of all sources and given an admixture...........the only issue to me is the naming of these groups and where they reside in the world.

example, someone like HDGP calls tuscan ..north-italian and incorporates together the central and north italy....others have a seperate ( MDLP, eurogenes ) one for north italy and tuscan for central italy