Weird output of DESeq2 when using single one sample from one group agains the other group (6 samples), really need some help
1
0
Entering edit mode
5 months ago
FantasticAI ▴ 60

Hi

I'm currently working on RNA-seq analysis, and my goal is to get differentially expressed genes using DESeq2.

I have 2 groups, A and B, and each group contains 6 samples. After using DESeq2, I got 1405 DEGs in total. I used p-adj < 0.05 as threshold. Code as following:

dds <- DESeq(dds)
res = results(dds, cooksCutoff=TRUE,
                  contrast=c("condition", "group A", "group B"),
                  independentFiltering=TRUE)

However, when I tried to use one single sample from group A against group B (6 samples), the number of DEGs is very small:

dds.subset <- DESeq(dds.subset)
res = results(dds.subset, cooksCutoff=TRUE,
                  contrast=c("condition", "group A", "group B"),
                  independentFiltering=TRUE)
sample 1 from group A vs group B: 11 DEGs
sample 2 from group A vs group B: 245 DEGs
sample 3 from group A vs group B: 35 DEGs
sample 4 from group A vs group B: 21 DEGs
sample 5 from group A vs group B: 7 DEGs
sample 6 from group A vs group B: 20 DEGs

Although using one single sample from group A against group B (6 samples) will lead to different number of DEGs comparing to group A(6 samples) vs group B (6 samples), but the numbers I got is too few and I believe something went wrong.

The possible reason I guess is that using one single sample, the estimated size factor and dispersion are changed, this will lead to another different fitted regression model. But this still cannot explain why the number of DEGs is small.

Another possible reason I guess is that using one sample against the other group (6 samples), DESeq2 combines 6 samples together to one large sample so that the raw count for each gene is larger than before. This lead to a comparison of small count vs large count, so in the end, only the genes which has really high raw count in the single one sample from group A can be recognized as DEGs.

Any help would be appreciated, thanks in advance!

WaldTest DESeq2 NegativeBinomial • 526 views
ADD COMMENT
0
Entering edit mode
  1. Why do you compare samples in groupA vs B separately?
  2. Try some other packages, such as edgeR, for no-replicate condition.
ADD REPLY
1
Entering edit mode
5 months ago

Cross posted and answered by the DESeq2 developer: https://support.bioconductor.org/p/9141742/

ADD COMMENT

Login before adding your answer.

Traffic: 2125 users visited in the last hour
Help About
FAQ
Access RSS
API
Stats

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6