Who can see your viewing activity?
Cluster number could be highly variable depending on your method. Which method are you using?
Can you look at boundaries between regions where there might be interesting receptor-ligand interactions?
How do you account for different cell cycle phases when predicting differentiation trajectories?