AUCell scores are weird don't make sense

[dmalzl]

Hi,

I am new to using SCENIC and am quite puzzled over the meaning of the AUCell scores especially their magnitude. From the description of the score it seems that it is more or less an AUC value for a custom ROC curve. This implies that the scores should be between 0 and 1 which is also indicated by all plots in the R tutorials. However, running the pySCENIC pipeline I get all sorts of non 0 - 1 values with also many negatives. This also makes it hard to compare the scores between groups. Is there any advise on why this is? Did I miss something in the implementation or is this a bug?

Thanks in advance

[dmalzl]

I did a bit of forensic digging in the code base and found that this apparent "weirdness" is due to the use of weights (i.e. noweights = False) which especially for log normalized data etc. results in negative values due to the expression values being used as weights (or something like this). So if anybody wants the expected 0 - 1 value range set noweights = True