Low fraction reads in cell metric in snRNA-seq data

Question

For some of our snRNA-seq samples we are finding low fraction of reads detected in single-nuclei rna-seq samples from cellranger, while the other metrics are perfect.

While I understand this could be caused by library preparation and abundance of ambient RNA. Just wondering if this behavior is expected for single-nuclei datasets. My confusion is specifically coming from the fact that 10X website also has such a sample that has low number of fraction reads in cells.

https://cf.10xgenomics.com/samples/cell-exp/7.0.0/5k_mouse_brain_CNIK_3pv3/5k_mouse_brain_CNIK_3pv3_web_summary.html

Would love to know if this behavior is expected or seen often who are more experienced in snRNA-seq. The metric documentation page is here

Here is snapshot from 10X report

Answer 1

Yes, "low fraction reads in cells" is more common with snRNA-seq from frozen samples, I wouldn't worry about it by itself (you're just throwing away some sequencing data). You should still check though that the other QC metrics are good. In particular, sometimes such samples have an unclear seperation of cells from empty droplets on the UMI rank plot and the consequent estimated cell count is off, requiring alternative approaches to distinguish cellular droplets from ambient RNA etc.