Hello, I'm working with DESeq2, and when look my normalised mean expression in the Case vs Control the date doesn't match with the log2 fold change. Example:

gene baseMeanCase baseMeanControl   log2FoldChange  lfcSE   stat    pvalue    padj
TMTC1   463.67    379.85            3.1043          0.5566  5.5766  2.45E-08  0.0005

In my design I added : design = ~ sex + year + condition

In other hand, I performed my PCAs with limma::removebatcheffect to modify my ntd :

 gene      ntdCase    ntdControl
 TMTC1     9.81       6.74

And those values are in accordance with log2FoldChange.

I would like to have the expression values (not transformed) that could correspond to the ntd value.

Thank you for your time :)

For some genes (large counts, low dispersion) the naive logFC is close or identical to the moderated one from DESeq2. For others it is not due to the moderation, which is a key point of the method. You cannot 'extract values' to manually reproduce that since it is all model-based. This has been asked before, please find via google. Your best choice for downstream is probably the vst with blind=FALSE.