Hi! I have rna-seq data for samples that were exposed to environmental stressors across two timepoints in a fully factorial design (so, let's say my sample treatments are CC, CS, SC, and SS, where c is control, s is a stressor, and the first letter is the treatment of the first exposure and the second letter is the treatment of the second exposure). I have used DESeq2 to identify differentially expressed genes within pairwise comparisons. Based on the DESeq manual and other threads, I've applied a lfc shrinkage estimator to be able to compare my LFCs across pairwise comparisons - is that correct? or should I be using the raw LFC values?

Thank you!

Try the vignette which has a lot of information on these type of questions.

You are on the right track. In DESeq2, raw log2 fold changes are often noisy and can be misleading, especially for genes with low counts. Applying LFC shrinkage makes the effect size estimates more stable and easier to compare across pairwise contrasts in a fully factorial design like CC, CS, SC, and SS. The shrunken LFCs are therefore better suited for interpretation and biological comparison, while the raw LFCs should mainly be used internally for statistical testing. Being consistent with the same shrinkage method across all contrasts is important to ensure meaningful comparisons. @ wheelie life