One thing I noticed, however is that it doesn't work very well for diffusing cold values. In my case, I want to diffuse underexpressed genes. It performs as expected, but the side panel is extremely misleading since ranking is all done based on positive values. In my case, it's the negative values that matter. This is probably a very common case for expression data. I'm wondering if there should be some sort of flag in the algorithm (or the panel?) that re-ranks based on cold (i.e. negative) values rather than only the positive values.
Via @scootermorris