Just got my results back from AncestryDNA. I’ve done autosomal ancestry tests with 23andMe and Family Tree DNA, in addition to analyzing the raw data myself. When I first got the results the Y and mtDNA results were totally expected; R1a1a and U2b respectively. Those are typical South Asian lineages. But what surprised me was how East Asian I turned out to be. This is really obviously in ADMIXTURE and PCA, but curiously 23andMe, even in speculative mode above, gives a relatively low fraction. Below 10%. You can see in the AncestryDNA results that in fact I’m a lot more East Asian than that, as well as a Melanesian-like element. I suspect that the latter is due to Austro-Asiatic admixture, in particular from my mother. 
Family Tree DNA gives similar proportions.
The differences don’t invalidate the robustness of admixture analyses. It’s just that these results are always carved around the joints and reifications which are human-digestible, and humans are the ones who pick the specific parameter values and data inputs around which the models are constructed. The companies have different reference populations, and these reference populations are the data inputs around which individuals are constructed as linear combinations. 23andMe uses a sophisticated system that looks at haplotype blocks, but it seems quite clear that finely admixed people are sometimes difficult for them to resolve (because of the uniqueness of their blocks in comparison to their reference populations).
Comments are closed.