Closed grst closed 2 years ago
I think it is relevant for both the simulator and training of second-generation methods. We could maybe implement a general module for preprocessing?
Good idea. In that case it should probably go to omnideconv
?
Good idea. In that case it should probably go to
omnideconv
?
Do you mean omnideconv\omnideconv
? I would say so, unless we do not envision something like omnideconv\preprocessing
Do you mean omnideconv\omnideconv
yes. I don't think we need another package.
Not sure in which repo this fits best as it makes sense essentially everywhere where signatures are generated. In particular also relevant for @constantin-zackl's future signature curation efforts.
The idea is to distinguish between cells of which we are (relatively) confident the annotation is correct, and less confident ones.
A solution could be a graph-based appraoch as discussed in https://github.com/omnideconv/k-mer-deconvolution/issues/2