Does it make sense to analyze snRNA-seq data with low nUMI per gene?

Question

[fifty_fifty]

I have the results from cellranger with high background (~80 000 estimated cells) and ~8000 mean reads per cell (which is usually 10 times higher than that). However, all QC parameters of cellranger are ok.

I am trying to manually filter out empty droplets. There is a low number of counts per gene and very high percentage of mitochondrial genes per cell:

I filtered nuclei based on low/high genes and counts as well as I removed all nuclei that had more than 3% of mitochondrial genes. Now I have almost 5000 nuclei. However, the counts are very low, so nuclei do not cluster very well.

Does it mean this dataset should be discarded due to poor quality?

Answer 1

[ahmad mousavi]

Hi

Please clarify you have performed scRNA or single nuclei, if the second you may not get too many MT genes.

But about the low count per cell, based on what you show here there are not too much gene & reads. There is positive correlation between read count per cell and number of genes, the samples with 20K-30K reads has the best result for us although in some cases even lower than 20K has some information.

But such issues is depended to budget and availability of samples ....

Take a look at this paper

Missing data and technical variability in single-cell RNA-sequencing experiments