Gene expression normalization
Closed this issue · 2 comments
joan-yanqiong commented
Hi, I want to use the tool with scRNAseq data. Is normalization needed? Or is UMI fine?
earmingol commented
Hi!
It depends on the aggregation method you use for summarizing the expression within a cluster. If you are using an aggregation of the fraction of cells expressing a gene, UMI is fine. However, if you plan to compute the average expression within a cell type, or any other, I recommend using log1p(CPM). The same applies for using distinct communication scores: if you are using continuous ones it's better using log1p(CPM). Also, it is important that your cell type annotation is done after batch-effect correction if you have multiple samples.
I hope this helps!
joan-yanqiong commented
Hi, thank you for the help!