Question

How do 'low counts' change

0

Entering edit mode

Jessica ▴ 10

@jessica-24203

Last seen 4.2 years ago

Hello, I found out when I used DESeq2 to analyze the data, the low counts is very high. Most of padj is 'NA'.

out of 31557 with nonzero total read count
adjusted p-value < 0.05
LFC > 0 (up)       : 4, 0.013%
LFC < 0 (down)     : 7, 0.022%
outliers [1]       : 0, 0%
low counts [2]     : 22025, 70%
(mean count < 165)

My genes' expression are very low, this can lead to a lot of lost information. Is there any way to solve it?

dds <- DESeq(dds)
res_table <- results(dds,contrast = c("condition","trt","untrt"),alpha = 0.05)

deseq2 • 891 views

ADD COMMENT • link updated 4.2 years ago by Michael Love 43k • written 4.2 years ago by Jessica ▴ 10

score 0 · Answer 1 · 2020-09-23

0

Entering edit mode

Michael Love 43k

@mikelove

Last seen 5 days ago

United States

You can set independentFiltering=FALSE, but you will obtain fewer DE genes. Here the low count filtering is helping to reduce multiple testing burden (read more about this in the vignette or the DESeq2 paper).

ADD COMMENT • link 4.2 years ago Michael Love 43k

0

Entering edit mode

I tried to set independentFiltering=FALSE, as you say I got fewer DE genes. The result not good.

As I mentioned before, some of my genes' expression are high, most of them are low. Is there any way to analyze such data? Otherwise I would lose a lot of data.

ADD REPLY • link 4.2 years ago Jessica ▴ 10

0

Entering edit mode

I don’t have any further suggestions from what you tried initially.

ADD REPLY • link 4.2 years ago Michael Love 43k

0

Entering edit mode

One more question, how to define low counts? It means why the low counts : 22025, 70% (mean count<165), rather than mean count<1 or something else. How is it calculated?