QC cut off for snRNAseq data

Question

[AFP3]

Hi all,

Apologies if this is a silly question as I am very early on in my bioinformatics journey! I have been playing with an snRNAseq dataset in Seurat and I am currently performing QC on the dataset. Alot of the profiles look like those shown below 1.

So my understanding of the QC step is to remove of situations when we have multiple cells in a well/just debris in a well. From the examples I have seen, the "tail" on nFeature_RNA tends to be alot thinner than the data I have, and so I am slightly unsure of the course of action! Would a suitable cutoff here be something like 600< nFeature <3000, percent.mt<5. percent.rib <2. Or am I misunderstanding what those values above nFeature_RNA > 3000 are?

Many thanks in advance for the help!

Answer 1

[rpolicastro]

Instead of semi-arbitrarily picking thresholds, I tend to gravitate towards the MAD strategy outlined here. It's more easily reproduced and scales better once you start having more samples.