Hi everyone,
First time poster here - tried to look for the answer but I do not seem to find exactly what I am looking for.
I have a question re log2FC shrinking as part of my DEse2 pipeline. I cannot understand which type of shrinkage I should be using.
Data is from totalRNA extracted from a tissue + vs - KO of a gene of interest. Aligned to genome with STAR, deduplicated, and then gene counts extracted with featureCounts. The comparison is a simple 4 WT replicates vs 4 KO replicates. No complex experimental design.
I know in advance that the knock out affects a lot of genes which are highly expressed. We normally would shrink the log2FC using apeglm but this time, if I observe the MA plot after apeglm, the results look very weird.
These are the unshrunken, the apeglm shrunken, and the normal shrunken MA plots
One can see here, although not easily, that there are some significant blue dots almost on the 0 axis, in the apeglm shrunken data. Something I could also observe in my volcano plots.
From what I understand the effect of shrinking should be, I should be trusting in this case the normal one, more than the apeglm? Considering that despite the apeglm shrinking, I still get some log2FC for lowly expressed genes to be relatively high (ie +3 log2FC)
Additionally, although it is sometimes only a moderate effect, the genes significantly regulated have bigger log2FC (in absolute value) when I do apeglm, even those more highly expressed.
Is there any way to confidently say which method is correct? Is the shrinkage affected by the fact that I do not have a "balanced" amount of genes changing in positive vs negative?
Thanks very much for your help.
Chiara
Here below the code I used for the plots after running the DESeq for making the plots
lfc_shrink <- lfcShrink(dds, coef=paste("condition", treatment, "vs", control, sep = "_"), type="apeglm")
lfc_shrink_n <- lfcShrink(dds, coef=paste("condition", treatment, "vs", control, sep = "_"), type="normal")
FDR = 0.1
DESeq2::plotMA(res, alpha = FDR, main = paste0("Unshrunken MA-Plot (FDR = ", FDR, ")"), ylim = c(-5,5))
DESeq2::plotMA(lfc_shrink, alpha = FDR, main = paste0("apeglm MA-Plot (FDR = ", FDR, ")"), ylim = c(-5,5))
DESeq2::plotMA(lfc_shrink_n, alpha = FDR, main = paste0("normal shr MA-Plot (FDR = ", FDR, ")"), ylim = c(-5,5))