Question

Strange lfcShrink plot when running DESeq2 workflow

0

Entering edit mode

Ben ▴ 50

@ben-17772

Last seen 3 months ago

United States

Hi everybody,

I am producing right now a strange plot while running the DESeq2 workflow when running the command lfcShrink on the DESeqDataSet object (after running the DESeq function).

I assume right now that there is an error within my workflow but I can not figure out what the problem. In a nutshell this is what I was running on the dataset:

import of dataset using tximport (RSEM result files),
creation of design matrix using limma,
creation of DESeqDataSet using the tximport object and the design matrix,
pre-filtering for low-expressed genes,
running DESeq function on DESeqDataSet,
creation of contrast matrix,
running lfcShrink on the DESeq processed DESeqDataSet.

I can update the code of course but maybe someone has some ideas just based on the plot itself.

Thanks much in advance!

left plot showing results(object=dds,...), right plot showing lfcShrink(dds_calc,...)

Both plots plotted using DESeq2::plotMA(...).

DESeq2 rnaseq rna-seq deseq2 deseq • 1.2k views

ADD COMMENT • link 5.5 years ago Ben ▴ 50

score 0 · Answer 1 · 2018-10-12

0

Entering edit mode

Michael Love 41k

@mikelove

Last seen 14 hours ago

United States

I've seen this before when users are plotting the intercept.

dds <- makeExampleDESeqDataSet()
dds <- DESeq(dds)
res <- lfcShrink(dds, "Intercept")

ADD COMMENT • link 5.5 years ago Michael Love 41k

0

Entering edit mode

Hey... Thanks for your quick response...

But no... I do not plot the intercept... (below part of my output)

> show(dds_design)
   DF_4 DF_9 DHF_4 DHF_9
1     0    1     0     0
2     1    0     0     0
3     0    1     0     0
4     1    0     0     0
5     0    1     0     0
6     0    0     1     0
7     0    0     0     1
8     0    0     0     1
9     0    0     1     0
10    0    0     0     1
11    0    0     0     1
> resultsNames(dds_calc)
[1] "DF_4"  "DF_9"  "DHF_4" "DHF_9"
> resLFC=lfcShrink(dds_calc,coef=2,type="normal")
> resNoLFC=results(object=dds_calc,contrast=contMatrix[,2])
> graphics::par(mfrow=c(1,2))
> DESeq2::plotMA(object=resNoLFC)
> DESeq2::plotMA(object=resLFC)

ADD REPLY • link 5.5 years ago Ben ▴ 50

0

Entering edit mode

This design should not be used with LFC shrinkage. The coefficients represent the base level for each group, and you don't want to shrink these to zero. It's essentially shrinking the intercept (each group has its own intercept). You should only use LFC shrinkage on coefficients representing differences between groups.

ADD REPLY • link 5.5 years ago Michael Love 41k