Hi everyone,

I Did DESseq2 differential expression analysis for 180 samples from TCGA using HTSEQ raw counts. After following all the steps and normalizing for two factors i got the expressed data with lof2FC, pval and padj. Then i did log-fold shrinkage using the apeglm package and im getting differentialy-expressed genes with significant p-values but very small logfold changes, for example 0.0006, which should be the result of many samples i think. When i make a volcano plot i see genes overlapping at the center of it, something that does not happen when differential expression without log-fold shrinkage is done. Is there any statistical criteria for selecting those genes, for example, starting from log2FC 0.2, or 0.5? I know you can chose for highly-expressed genes, but my focus is enzymes which do not show a lot of change.

Thank You

The lfcThreshold (which can be specified in results or in lfcShrink) is up to you, depending on what you feel is a biologically relevant LFC. You might look at the MA plot of the shrunken LFC and decide based on that distribution.

Thank you Michael for the quick reply.

Actually the shrinked lfc MA plot is something like this

https://www.dropbox.com/s/hun8s44twzxw3g2/Shrinked%20lfc%20plot.tiff?dl=0

But i think i got the answer.