I just completed my first very large multi-timepoint, multi-mouse, multi-protocol, multi-condition analysis in scanpy. LOVED working in scanpy, thank you to the developers for this amazing resource.
Looking for suggestions on taking the next analysis step. I have some interesting clusters, and some interesting gene hits. I’ve done GO and GSEA (shoutout to gseapy for making it easy to do this in python). I’d like to: discover (in a more sophisticated way than GSEA/GO) what sort of networks are enriched, and discover new networks. E.g., JAK/STAT signaling is elevated at an early timepoint, but suppressed later, or that the entire DE between condition 1 and 2 in cell type A is best explained by upregulation of TFs X, Y and downregulation of Z.
Some R packages that seem to do this include PROGENy/DoRothEA (Saez lab). I’ve found a few that seem spiritually similar in the python universe as well, but I’ve not seen anyone use these except the authors. For example, CellOracle works with scanpy… but it appears you have to start your analysis over in a new environment, and it was optimized for scseq+chipseq. That’s a problem because I’m now sort of married to BBKNN or scanorama for batch correction. Today I came across DrivAER which seems to meet some of my needs, installed it today and started to work with it. But I’m just wondering if there are any packages designed with scanpy in mind that achieve some of these goals of identifying significant TF / signaling pathways. Appreciate any input!