Any ideas on how one can simulate pathways in a scRNA-count table?

I want to test the performance of a method on whether it can pick up pathways-related signals from the scRNA-count table.

It's like having pathway-based meta-labels for gene sets in the scRNA-count table.

So, if a pathway is significant in the entire dataset, the genes related to that pathway should be differentially expressed and have some distinguished signal relative to other gene sets.

I thought of combining small scRNA datasets simulated using splatter. In this case, each simulated dataset will correspond to a pathway. But does it make sense?