I'm performing a differential expression analysis running DESeq2 and came across a question:
When I'm getting significant DEGs, should I subset my DEGs object using 'log2FoldChange' column from the results() output or from the lfcShrink() output?
In some of my analyses, this column may present quite different values after running lfcShrink().
I appreciate any help.
In some of my analyses, this column may present quite different values after running lfcShrink().
That's the point of the method. Essentially, the lfcShrink function tries to "correct" logFCs with large standard errors. That is essentially when there is little evidence that the observed logFC is true, then it corrects it ("shrinks it") towards zero. That is especially true for genes with low counts. The vignette recommends to use the shrunken logFCs for downstream analysis such as ranking purposes.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Thank you so much for your response! I'm a bit confused exactly on that... So if I want to select some genes based on logFC, I should rank them and then select those above a cutoff (i.e. logFC > 1.5)?
Selecting based on a value has nothing to do with ranking. Ranking is a potential downstream analysis, for example together with gene set enrichment analysis. If you feel like subsetting by fold change then just do that, it has nothing to do with ranking. If you want to do subsetting then lfcShrink would be better than results output indeed.