Strange faststructure results - K5 plot only has 3 colours/clusters.

[miranda...@gmail.com]

Hi all,

I've run faststructure before and with no issues, but now something a bit strange is happening.

K = 4 and 5 are selected (for likelihood and maximising structure) but when I check the plots, K = 5 only shows 3 colours (attached plot). i.e. there is very low membership for the other 2 clusters in the meanQ. 

In adegenet and other pop gen analyses 3 clusters are identified so this makes more sense.  The plot for K = 3 is very similar. 

Does anyone have an idea what is going on here?

[Francisco Pina Martins]

Dear Miranda,

From my experience, fastStructure's "choosek.py" is not particularly reliable. I have witnessed what you describe (choosek.py overestimating the best value of K compared to other approaches) occurring on multiple datasets. I'm not sure why this happens, but I tend to disregard the results of "choosek.py" anyway as it frequently seems to overestimate the best K value.

Hope it helps,

Francisco