Hi all,

I have two cell types, (A) a cell line and (B) a primary cell. I then added the same stimulus on them and sequenced their RNA. My goal is to show that even though their not the same cells, they react similar to the stimulus. I have ploted the PCA of the expression matrix where the untrated (A and C) and the stimulated (B and D) are plotted seperately. Is there a statistical way to asses the similarities of the differentially expressed datasets (DEG1 and DEG2) and plot them in a selfexplanatory way? I've used DESeq2 for my analysis.

Thanks!

Sounds like a job for gene set enrichment analysis, e.g. via http://bioconductor.org/packages/release/bioc/html/fgsea.html

That will check if a given geneset, e.g. DEG1 or DEG2 shows evidence to be consistently up- or downregulated in a given comparison (as a whole set, not individually per gene). You can take DEG1 as a gene set and run against the C=>D comparison and vice versa. If the effects (so the DEGs) are similar then the gene set should come out as signifcant. Basically it asks whether the rank distribution (you rank all genes e.g. by signed pvalue or fold change) for that gene set is more extreme than one would expect by change, and that is achieved by creating random permutations of the gene set to compare against. See also fgsea usage question - what gene set to use?

For the ranking I like the shrunken logFCs from lfcShrink. Alternatively, signed -log10(pvalue) (not padj because that has many ties) makes sense. Signed means it gets minus if the logFC is negative.

Is this what you need?