TF all down-regulated in a cluster after rankby group.

[wariobrega]

Hello,

First, congratulations for the new version of the package, it is comprehensible, accessible, and easy to use :).

I have a small question on the decoupler ulm and rankby_group methods.

In brief, I am performing differential transcription factor analysis on a scRNA-seaq dataset. I am specifically interested to TF activity differences between two clusters of interest.

I have my pre-computed GRN that I derived from a GRN inference method and I used it to compute TF activity with decoupler using the univariate linear model.

After that, I performed the rankby group method, using both 'rest' and one of the two cluster of inference as the reference pool.

What I notice is that, for one of these two clusters, there are no up-regulated TFs. I found this weird, since I see that the expression patterns for some of the targets of these TFs are either very high or very low.

I was wondering if such cases are expected and if so, what could be some possible explanation.

Another thing I was wondering, if the variance of the per-cluster tf activity is too high (and that might be the cause of this issue), would it make sense to normalize the TF activity values before doing the differential tf activity with rankby group? Have you ever tested it ?

Thanks for your time,

All best,

Daniele