Question

DESeq2 lfcshrink generate 0 svalues

0

Entering edit mode

corentin.lyon • 0

@corentinlyon-22056

Last seen 4.6 years ago

I run DESeq2 using lfcShrink to get LFC estimates:

resLFC <- lfcShrink(dds, coef="condition_T_vs_O", type="apeglm",lfcThreshold = 1,svalue = TRUE)

I then want to plot a volcanoplot using svalues:

ggplot(as.data.frame(resLFC), aes(x=log2FoldChange, y=-log10(svalue),color=svalue<0.005)) +xlab("log2(Fold Change)")+ylab("-log10(svalue)")+theme(legend.position="none")+
  geom_point()+theme_bw()+scale_color_manual(values=c("#1d1d0a","#9a3348"))

enter image description here

Why in O condition, almost all the svalues are set to 0 for negative LFC genes? And not for positive LFC genes?

deseq2 • 848 views

ADD COMMENT • link updated 4.6 years ago by Michael Love 41k • written 4.6 years ago by corentin.lyon • 0

score 2 · Accepted Answer · 2019-10-04

2

Entering edit mode

Michael Love 41k

@mikelove

Last seen 1 hour ago

United States

I tend to use adj p-values/s-values just for setting a threshold and then creating a FDR/FSR bounded set. And then I prefer to look at MA (difference over signal strength) plots.

You may explore these genes with plotCounts but basically, these are nowhere near having a false sign.

ADD COMMENT • link 4.6 years ago Michael Love 41k

0

Entering edit mode

Ok this makes sense, Last thing I am not sure to understand, to have a FDR of 1% should I use a treshold on 0.01 padj or 0.01 svalue?