Why is lfcshrink() used for RNA/Ribo and not for TE?

[evaesquinas]

Hello,

Although I am quite familiar with DESeq2, I haven't ever used lfcshrink() and I wonder why:

• You used it (?)
• You used it to obtain the log2FC for RNA and Ribo (but not for TE).

This is part of the code taken from the script where you add it or not.

TE

> ddsMat = DESeq(ddsMat)
#Obtain fold changes for TE:
> res = results(ddsMat, name=“Condition2.SeqTypeRIBO”)

RNA:

> ddsMat_rna = DESeq(ddsMat_rna)
> res_rna = results(ddsMat_rna, name="Condition_2_vs_1")
> res_rna = lfcShrink(ddsMat_rna,name="Condition_2_vs_1",res=res_rna)

RIBO/RPF:

> ddsMat_ribo = DESeq(ddsMat_ribo)
> res_ribo = results(ddsMat_ribo,name="Condition_2_vs_1")
> res_ribo =lfcShrink(ddsMat_ribo,name="Condition_2_vs_1"),res=res_ribo)

Thanks very much in advance

Regards

[thyagoleal]

@evaesquinas I believe you should the lfcshrunk log2FC every time you would consider the log2FC to rank genes or categorize as differentially expressed. Without the shrinkage, genes with high and low expression sometimes have inflated log2FC that can be irreproducible. I don't know why they haven't used for the TE, but I assume is because "A DTEG is determined based on significant change in TE (FDR < 0.05)" (Chotani et al., 2019).