I have been interested in using AUCell as a way to determine significantly enriched pathways in my single cell spatial transcriptomics data set (CosMx) and wanted some clarifications on the results.

From my understanding, AUCell just assigns a signature score to every single cell in your dataset and it does not do anything like test for significantly enriched scores. In my dataset, I have many different samples that can be partitioned based on some feature (ex. type 1 vs type 2). Is it reasonable to calculate the proportion of cells in each sample that pass a threshold for a given signature and then compare the distribution of proportions across groups (ex. a t-test or wilcox test)? I would like to say something like "type 1 had significantly more cells that were enriched in a TGFb signature versus type 2."

Thank you in advance!