Parameter of dispersion to use

I have a quick question, first here is an example:
These are two paramaters, let's say genes A and B and their level of expression in 9 samples.
A: 5 5 7 2 2 8 6 5 3 var= 4,44
B: 0 0 0 0 0 0 3 3 3 var= 2,25
Usually people in the field look for the highly variable genes (HVG), in this case it is the gene A since it's var > var gene B. But for me, the gene B is more interesting biologically since it is expressed in a few samples and not in the others. So my question is, what is the best dispersion paramater to use if I want to see the same thing (genes showing this trend of dispersion as in gene B) in a bigger dataset with a lot of samples and many thousand genes?
Thank's for your help