I have a query regarding what's the right approach for doing DE analysis for scrnaseq data where we have ctrl vs treatment samples. The data was integrated with atlas and DE was performed between ctrl and treatment cells in each cluster separately. So now we have DEGs per cluster and I wish to perform ORA analysis. The way I can do this is as follows
- Take DEGs from all the clusters, remove the duplicate gene IDs and then perform ORA. The background here could be union of genes expressed in both ctrl and treatment.
- Take DEGs from a cluster and then only the genes expressed in that clusters as background and then perform enrichment per cluster by looping over
- Take DEGs from all clusters. Set an empirical cutoffs like (expressed in atleast 100 cells) for background construction. And then perform ORA
Which one ise better?