Question

disagreement between log2FoldChange and lfcMLE in DESeq2

0

Entering edit mode

aec ▴ 90

@aec-9409

Last seen 5.4 years ago

Dear all,

I am using DESeq2 for a design ~PAIR+GROUP and I obtained for geneX this result:

	geneX
baseMean	62.2087194
log2FoldChange	-0.33167142
lfcMLE	-2.32040851
lfcSE	0.07892334
stat	-4.20245054
pvalue	2.64E-05
padj	0.06846016

The normalized counts by DESeq2 are these:

	geneX
groupB	17.51
groupB	13.94
groupB	31.76
groupA	82.32
groupA	81.39
groupA	146.33

Why is it that the log2FoldChange is so different from the lfcMLE? The gene is clearly down-regulated in groupB by a factor>2 and this is not reflected in the log2FoldChange.

deseq2 log2fc shrinkage • 1.8k views

ADD COMMENT • link updated 7.9 years ago by Michael Love 43k • written 7.9 years ago by aec ▴ 90

score 0 · Answer 1 · 2017-12-14

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 1 day ago

United States

You may be using an older version of DESeq2? The latest version (1.18) doesn't perform LFC shrinkage automatically, but only after running the command lfcShrink(). In addition, the latest version has new estimators which tend to shrink less than the originally proposed shrinkage estimator in DESeq2. Shrinkage is useful for ranking and visualization, but on some datasets, with a particular distribution of underlying LFCs, we found the original shrinkage estimator to be too strong. This could be the case here as well.

ADD COMMENT • link 7.9 years ago Michael Love 43k

0

Entering edit mode

Mike I am using version DESeq2_1.10.1, is it too old?

ADD REPLY • link 7.9 years ago aec ▴ 90

0

Entering edit mode

That version is from 2 years ago. I'd recommend -- if you're on a laptop, or have permissions to do so -- download the latest version of R and update your Bioconductor packages (unless you need to stick with the 2 year old version for some reason, ... e.g. an ongoing project where you need to continue working with an old set of packages).

ADD REPLY • link 7.9 years ago Michael Love 43k