to consider only intronic reads in cell ranger pipeline

Question

[Dataminer]

Hello community,

I am processing some single nuclei RNA seq data and unfortunately while performing the experiment there was high leakage of mRNA and that leads to ambient RNA all over.

I thought may be considering only intronic reads while mapping would circumvent this problem, but I am not sure how to do this.

Has anyone encountered this problem?if yes, how did you fix it or what would be your suggestion other than redoing the experiment.

Many thanks in advance.

Answer 1

[rpolicastro]

Extensive work for RNA-velocity (Manno et al., 2018, Bergen et al., 2020) has shown that most intronic reads are derived from aberrant priming of nascent transcripts, which in most cases must be from the nuclear RNA fraction. The software designed to count features for RNA-velocity will count reads based on whether they were likely derived from a spliced or unspliced (or ambiguous) transcript. This review nicely covers the workflows for doing this sort of feature counting. As for the efficacy of only using unspliced reads I'm not sure, so I'll let someone else answer if they know.