Empty clusters

[wedge015]

Hello,

On two new installs of fastStructure (mine on 3/11/21 and a colleague's on 3/15/21) on two different machines we ran into an issue where fastStructure was not utilizing all of the allowed clusters. For example, when I ran K=5, cluster 1 and 3 had no individuals assigned to those clusters. At K=6, only 4 genetic clusters were used. This continued at higher values of K, never utilizing more than 4 clusters. My colleague ran into the same issue independently with an unrelated data set.

Thanks in advance for any help or advice!

[RvV1979]

I can confirm this. Some of my recent analyses result in at least one of the clusters having a constant estimated assignment fraction among all individuals (0.000001).

I personally consider this is a fatal flaw and will no longer use faststructure for my work.