Scanpy: Is it possible to test for bimodality in scanpy?

[erikadudki]

Is there a way to test if particular genes of a single-cell-RNA-seq-dataset show bimodal distributions, in order to infer heterogeneity?
e.g. like in: https://www.nature.com/articles/nature13437

[erikadudki]

@Koncopd @ivirshup @flying-sheep @gokceneraslan Is there something in this direction implemented in scanpy? I would be happy to hear from you. If it exist already, I wouldn't have to implement it on my own :)

[LuckyMD]

Hi @erikadudki,

Sorry for the slow replies to this post. There is no test implemented for bimodality in scanpy at the moment. For modeling you could look into diffxpy, where you may be able to fit a gaussian mixture model to do model selection. Otherwise I guess statsmodels is the way forward for this in python.

[ColeWunderlich]

@erikadudki If you are willing to dabble in R, the scDD package seeks to address this problem specifically. It may also be worth looking at for inspiration if you want to pursue your own implementation in python.