Question

Identifying significantly differentially expressed genes across all the groups within a factor using DESeq2

0

Entering edit mode

gupta.anuj0608 • 0

@guptaanuj0608-15005

Last seen 6.8 years ago

I have a RNASeq data-set which consists of 4 different conditions (Control, Treatment1, Treatment2 & Treatment3) each with duplicate samples.

I would like to identify all genes that are significantly differentially expressed across all the 4 groups.

Does DESeq2 allow such assessment?

Thank you for your time and I look forward to your response.

deseq2 • 936 views

ADD COMMENT • link updated 6.8 years ago by marika.catapano • 0 • written 6.8 years ago by gupta.anuj0608 • 0

score 0 · Answer 1 · 2018-02-12

0

Entering edit mode

marika.catapano • 0

@marikacatapano-15006

Last seen 6.8 years ago

Of course you can!

in the function "results" just define your contrast.

e.g.: dds<-DESeq(dds)

Treatment1vs2 <- results(dds, contrast=c("group", "treatment1", "treatment2")) and so on for each combination

ADD COMMENT • link 6.8 years ago marika.catapano • 0

0

Entering edit mode

Hi Marika,

Thanks a lot for the reply. Sorry if I wasn't clear enough. I want to identify all genes that are significantly differentially expressed across all the 4 groups at once (i.e. reject null hypothesis that the genes have equal expression across the 4 groups). I don't want to make pair-wise comparisons. EBSeq multi-group analysis allows that but I would like to do that using DESeq2, if possible.

Thanks

ADD REPLY • link 6.8 years ago gupta.anuj0608 • 0

0

Entering edit mode

Take a look at the likelihood ratio test section of the vignette.

ADD REPLY • link 6.8 years ago Michael Love 43k

0

Entering edit mode

Hi Michael,

Thanks a lot for pointing that out. I am still not able to figure out what value should I provide to "reduced" argument. I only have one factor "cond" which consists of four groups. In all the examples, "~ 1" has been used as the only value to the "reduced" argument. Here is the code I am using:

> strcount.int)
int [1:26341, 1:8] 247 385 0 0 18 9 0 0 916 0 ...
- attr(*, "dimnames")=List of 2
..$ : chr [1:26341] "A1BG" "A1BG-AS1" "A1CF" "A2M" ...
..$ : chr [1:8] "expected_count_DA" "expected_count_DB" "expected_count_K15A" "expected_count_K15B" ...

> str(cond)
Factor w/ 4 levels "control","K15",..: 1 1 2 2 3 3 4 4

> dds <- DESeqDataSetFromMatrixcount.int, DataFrame(cond), ~ cond)

Thanks again!

ADD REPLY • link 6.8 years ago gupta.anuj0608 • 0

0

Entering edit mode

Yes. A design of ~condition compared to reduced of ~1 will test any differences in gene expression due to condition. Take a look at plotCounts, to visualize the results from this test.