Edge case where all perturbations are NP

[chriswzou]

Hi! Another issue - also happy to help fix this one if a quick PR is helpful. I was doing some testing of a single perturbation before starting a long run of them, and found that when perturbations all turn out to be NP, the RunMixscale function errors out.

It looks like it might be coming from this line of the code, where the cell types are drawn from the name of the first perturbation added to gv.list. In cases where gv.list is empty however—which I think happens when all the perturbations are NP—I get a subscript out of bounds error coming from trying to access gv.list[[1]].

One fix for this could be to derive the cell types from the splits variable.

[longmanz]

Hi @chriswzou ,
Thank you for catching this bug! You are definitely right that this is driven by the edge case when all perturbations are NP. We have commented out that line of code to avoid this.
Additionally, now we will throw a warning message to users if all the perturbations are NP and gv.list is empty.