what should I do if I get a compromised sample of scRNA-seq data?

Question

[wangjb702]

Hi everyone,

Recently, I got my snRNA-seq data. But after aligned to reference genome using cellranger, I got a wired barcode rank plot. what should I do next to correct this data? Thanks in advance

[dsull]

The knee plot is just one type of QC. I'd say just proceed with your analysis and see if you observe anything funky (analysis itself is a QC). You should also see if you can use a tool that removes doublets and ambient RNA, if that helps your analysis.

[ATpoint]

Agreed. I would just force the cells to 5000 or so, and then proceed with the normal QC. If there's a lot of trash "cells" in the dataset this will manifest by poor QC metrics such as number of detected genes and mitochondrial expression percentage (since mt genes usually are highly-expressed so they're always detected first), plus "trash" tends (in my experience) to cluster together with marker genes that make no biological sense. Just try a bit, see what works.