Dear all,

I am using DESeq2 to identify differentially expressed genes between two conditions. This is a summary of my code:

dds-tvsc <- DESeqDataSetFromMatrix(countData = cd-treated-vs-control_matrix, colData = experiment, design= ~treatment)
dds-tvsc_DESeq <-DESeq(dds-tvsc)
res <- lfcShrink(dds-tvsc_DESeq, coef =2, type = "apeglm", lfcThreshold=0.585)

As I have seen in the DESeq2 manual that it is now recommend to use the lfcShrink function. I am interested only in those genes with fold-change > 1.5 and for this reason I have used lfcThreshold command. Then, s-values appear in the output of the lfShrink function instead of p-adjusted values (FDR). As far as I understand, I have to check s-values to identify differentially expressed genes (as s-value is equivalent to FDR in the results function) ; however, I do not know how to set a s-value threshold. Is it correct to use 0.05 like with FDR from results?

Thanks a lot for your help

Jose

We’ve found that in practice, smaller nominal thresholds of s-values using lfcShrink give similar sized sets, eg .01 s-value for .05 adjusted p-value, although these are not the same set or error control. I find the “false sign or smaller than X” type of error control quite compelling for genomics. For more details, see the apeglm paper and the Stephens 2016 paper that is cited by apeglm.