Inspection of gene expression in scRNA-seq data

Question

I am running the data preprocessing pipeline for scRNA-seq data presented here.

3.8.6.1 Gene expression

In addition to removing cells with poor quality, it is usually a good idea to exclude genes where we suspect that technical artefacts may have skewed the results. Moreover, inspection of the gene expression profiles may provide insights about how the experimental procedures could be improved.

It is often instructive to consider the number of reads consumed by the top 50 expressed genes.

By analyzing my data I have the following most expressed 50 genes (evaluating the total number of transcripts across cells):

Original

Log-transformed

Since I'll be working on log-transformed data, I assumed that the distributions of the top 50 genes were not excessively skewed. However, as you can see in the first plot, there are some samples (cells) with very high number of transcripts for the top gene. Is it something I should worry about? What other insights can I get from these two figures?

Edit

Original analysis (figure attached) was:

The distributions are relatively flat indicating (but not guaranteeing!) good coverage of the full transcriptome of these cells. However, there are several spike-ins in the top 15 genes which suggests a greater dilution of the spike-ins may be preferrable if the experiment is to be repeated.

Answer 1

In an ideal world, ribosomal RNA (as seen in your top hit) should be excluded from samples prior to sequencing. Where this is not possible (i.e. in the data that have been presented to you), it would be a good idea to exclude ribosomal genes (and any other common contaminants) prior to doing further analysis (including normalisation). I believe this is the point the course instructors are trying to get across with that message box.

Eyeballing the results, you should also exclude any ERCC genes as well, as they are synthetic constructs and not from the target samples (although record their counts, as it may be informative for QC/normalisation purposes); they'll have odd distributions because of this.

ERCC sequences are typically added based on the concentration of DNA in a sample, rather than the number of cells. However with single-cell data (i.e. with known cell counts), it'd be possible to do a proper ERCC-based normalisation, assuming an equal amount were added for each cell. I suspect that knowing a bit more about the ERCC addition protocol would help in understanding why their distribution is so different.