Open Puriney opened 4 years ago
I consider it in this way. I think the interaction strength between a ligand-receptor pair is only related to two genes (proteins), but normalization will make this strength influenced by other genes due to the use of relative expression.
Hi,
https://github.com/wguo-research/scCancer/blob/3f9eb9e878531d588594a132f00d60d6d38506c8/R/scAnnotation.R#L1325-L1326
Here it uses the 'counts' slot , i.e., UMI, to estimate the strength of ligand-recept pairs. I was wondering if it is making more sense to use 'data' slot , i.e., normalized data.