question regarding QC of scRNA-seq data

Question

[Alex V. Gopanenko]

Hi all,

I wonder if there are any recommendations regarding the next questions related to scRNA-seq data QC processing:

1. The first one: If I have 3'-polyA scRNA seq data, how to deal with such genes as histones, lncRNAs? I know that some of the mRNAs of histones genes and lncRNAs could be polyadenylated, but I suggest that this is more the exception than the rule. I saw a lot of suggestions on how they can be incorporated into the libraries and thus being sequenced. But the question is what general recommendations to deal with it? Remove or ignore? If remove from consideration, all or maybe some gene set is already created for filtering?
2. The second one: I have a set of samples related to same cell type in different conditions and I also have a replicates of that samples. When I performed QC (e.g. with Seurat), and I saw that QC metrics for each sample differ significantly (nFeature, nCounts, mt.percent), should I apply different QC filtering parameters for each sample? Or is it better to try use the same parameters for all samples… I found the current best practices that are recommend (if I understood correctly) to choose first way. https://pubmed.ncbi.nlm.nih.gov/31217225/ But I'm afraid of misinterpretation.

I will appreciate it if someone could help and/or point me on how to resolve that questions

Thanks in advance,

All the best,

Alex

Answer 1

[rpolicastro]

1. Most people just ignore anything that isn't poly-adenylated. It's also common in certain analysis, such as Cell Ranger for 10X data, to only count protein coding genes.
2. You should apply different metric thresholds to each sample individually.

[Alex V. Gopanenko]

Thank you very much for your answer!

It is very helpful!

All the Best, Alex