The main message: the hundreds of thousands of genetic markers used by 23andMe (and other personal genomics companies, e.g. deCODEme) to infer genetic ancestry provide a much more detailed and accurate picture of the geographical origins of your genome.
No surprises here. The power of the type of genome-wide genotype data generated by 23andMe for ancestry prediction has been compellingly illustrated by a series of recent studies comparing genetic clustering with geographical origins. Here’s one example comparing a map of northern Finland with the genetic clustering of its population, based on the same type of data:
The geographical and genetic maps have the same colour scheme, so you can see immediately how astonishingly well geographical origins can be inferred from this type of genomic information. Similar maps have been generated for other isolated populations (e.g. Sardinia), and for fine-scale structure within broader regions such as Europe and East Asia.
None of the personal genomics companies has yet harnessed the full power of ancestry prediction algorithms, although 23andMe’s advanced global similarity tool appears to be the best attempt currently on the market (screen shot for their European substructure view below the fold.)
Blaine is clear that his interest in the results from 23andMe was primarily due to ancestry testing rather than disease risk prediction (he discusses his other results here). He’s not alone; as best I can tell, a substantial fraction of the personal genomics market is currently driven by a fascination with genetic ancestry. That’s a huge market: as Blaine noted in another recent post, one genetic genealogy company alone has sold over half a million testing kits.
23andMe’s advanced global similarity tool, Europe view: