Sunday, March 4, 2012
ADMIXTURE analysis of West Eurasia - K=13 run
OK, my worst fears about the K=13 run did not come true. I was expecting ADMIXTURE to come up with a nonsense cluster, or even get stuck. But the K=13 ended up being a fairly smooth and informative run, and almost as quick as the K=12. The only problem is that the new cluster to appear is centered on samples from East Central Asia, which is something of a disappointment, considering that this analysis is designed to study West Eurasian genetic substructures. However, I suppose this new cluster might be useful to those who are looking to characterize in more detail the nature of East Eurasian admixture in West Eurasia.
Key: Red = Caucasus, Orange = Mediterranean, Yellow = North Atlantic, Light Green = Sub-Saharan, Green = East African, Aqua Green = East Asian, Aqua Blue = East Central Asian, Light Blue = Southeast Asian, Blue = West Central Asian, Dark Blue = Southwest Asian, Purple = South Asian, Light Purple = Baltic, Dark Purple = North Asian. See spreadsheet for details. A table of Fst (genetic) distances between the clusters can be found here.
Update: I just tried a K=14, and after a quick look at the output, I'm going to stop the analysis, and not bother going up to K=15. The results appeared very shaky, featuring, for instance, a Burusho-centered cluster that spread out in a very choppy fashion across Western Eurasia. It seems that K=13 is the reasonable limit for this dataset, but also probably the optimal setting.