Normalisation for single nuclei RNA-seq
For single cell RNA-seq the typical workflow includes a normalisation step to account for variable sequencing depth. In Scanpy/Seurat, CPM (Counts per million) is a simple and common choice. We don’t normally normalize for gene length (like we would with full length transcript bulk RNA-seq, e.g. TPM) because with the 10X we only sequence the 3′ or 5′ end of the transcript. CPM is probably still not the best normalisation choice, but that’s not my question.
For single nuclei RNA-seq, we have pre-mRNA and so the quantification step includes counting of intronic reads. This would seem to introduce a gene length bias, so my intuition is that we should normalize the data differently and not use CPM?
• 25 views
Read more here: Source link