Hi,

at the moment I am analysing a simple dataset comparing two conditions using DESeq2. For DEG calling i usually go for padjust < 0,05 and foldchage of > 2. I am always wondering whether I should use for that the shrinked fold change or the fold change deseq2 is giving me straight after the analysis.

My code ist pretty straight forward using apeglm for shrinkage. A typical scatterplot I get after the analysis looks like this. I use the following code to extract the counts:

normalized_counts <- data.frame(counts(dds, normalized=TRUE))

Lots of gens which look like DEG in the plot are not due to shrined log2folchange. Is there any issue in my pipeline in your opinion?

Link to Scatterplot: https://ibb.co/HNF4nNZ

If you want to filter for LFC > 1, I would recommend to incorporate that into the test with:

res <- results(dds, lfcThreshold=1)

The shrunken LFC are good for ranking genes by effect size and for visualization. Note that you can also use lfcThreshold in lfcShrink(), but it will convert from adjusted p-value to s-value which has a different interpretation (see apeglm paper).