I have re-analyzed the data of Behar et al. (2010) using my Clusters Galore method. See my previous post on the HGDP panel for some technical details.
Here are the 47 clusters of the optimal mclust solution over the MDS representation retaining 26 dimensions:
Each row has the number of individuals who are mapped to each of the 47 clusters. Here are a few comments:
The discovery that Jewish populations can be subdivided into numerous clusters is not inconsistent with Behar et al. (2010) and their observation of the existence of three major clusters in Jewish populations. This is a difference of detail.
Most clusters strongly map to single populations; many populations with "tribal" traditions and high levels of sanguinity are split into multiple clusters, suggesting the existence of sub-structure in them. And, there are a few clusters spanning several populations, such as #1 (Balto-Slavic), #22 (Syrians-Jordanians-Lebanese), #29 (Ethiopians and Ethiopian Jews), #25 (Romanians and Hungarians), #31 (Iranian and Iraqi Jews).