Hello everyone,

I have just read the vignette about DESeq2 and I have a question about shrunken log fold changes and threshold test.

I use DESeq() and results() functions to get the gene set with absolute LFC > 1. I understood that the tests give me the results with LFC estimated by MLE.

However, some absolute LFC become lower than 1 after being shrunk by lfcShrunk function.

Is it possible and does it make sense testing for statistical significance of absolute LFC > 1 estimated by MAP (apeglm) ?

I apologize if the question has been asked before.

Thanks in advance,

Corentin

We do have inference for the posterior (shrunken) LFC but instead of the point null, we estimate posterior probability of false sign and use the s-value as defined by Stephens. It is analogous to the q-value or adjusted p-value but for bounding the rate of false sign. Just set svalue=TRUE. There is some more explanation in the DESeq2 and the apeglm vignettes.