Ways to analyze treatment changes in scRNAseq

Question

[ashahban]

Hi,

I am working with single cell RNA seq to determine gene expression changes in individual cells using two conditions: untreated and treated. I have merged the untreated and treated files into a single seurat object to generate featureplots of genes that we are interested in, but we are unable to detect strong expression of genes that we have validated in many different assays to be changed after treatment. I wanted to know if there was any way to visualize gene expression changes other than feature plots. for example, could we take the mean expression of our gene of interest in the untreated, and determine fold change in each individual treated cell relative to untreated average? Thanks so much in advance for the help!

[Friederike]

If you have genes that you expect to change, it's usually a good idea to check if the reason you're not detecting them in the single-cell setting is that the genes' transcripts aren't detected in many cells or at very low levels. For that, the jitter/violin/ridgeplots mentioned by rpolicastro are a good choice as well as abundance plots that try to capture both the number of cells a transcript is detected in as well as the abundance, i.e. magnitude of expression. Those plots can be done relatively easily with the dittoSeq package).

Answer 1

[rpolicastro]

You haven't given any information about your analysis, but if you want another way to visualize expression you can use a violin plot (often paired with a jitter plot),

If you want to find differentially expressed genes and you have biological replicates, a pseudobulk analysis is usually recommended. If you don't have biological replicates a wilcoxon rank-sum test between the two conditions can suffice.

As an alternative, semi-supervised dimension reduction such as MCML can let you extract features (genes) that separate your two conditions.