DESeq2 multiple group comparisons
2
0
Entering edit mode
5 months ago
Bhumi • 0

Hi, I have a column with a "group" variable in two datasets. In dataset#1 I have three types (A, B, C) within the group variable, and in dataset#2, I have four types (A, B, C, D). I want to perform Differential Expression for all three and four types, respectively, using DESeq2 such that I can assign up-regulated genes specific to each type. I tried doing the pairwise comparisons for dataset#1 as follows: A vs B B vs C C vs A And then took the up-regulated genes specific to type A, B, and C.

Now, I am wondering if instead I create a new variable and perform Differential Expression to be able to identify DE genes specific to each type as follows: A vs BandC B vs CandA C vs AandB

Is it the right way to go about it? Also, can I do so such a comparison with four types as well? Your suggestions are highly appreciated. Thanks!

R DESeq2 GeneExpression RNAseq • 411 views
2
Entering edit mode
5 months ago

Now, I am wondering if instead I create a new variable and perform Differential Expression to be able to identify DE genes specific to each type as follows: A vs BandC B vs CandA C vs AandB

Why not? You can create a new colData column, or, as the author says, (not exactly the same thing, but related)

You can use contrast with a list, e.g. list(c("A","B"), c("C","D")) and then listValues=c(1/2,-1/2)

https://support.bioconductor.org/p/9136689/

0
Entering edit mode

Thanks very much.

1
Entering edit mode
5 months ago

You can make such comparisons in a easier way (at least for me) by using edgeR as was mentioned in this post. Let me see if there is a similar way to perform this using DESeq2.

Best regards

0
Entering edit mode

Thanks very much. I think I could do that cause instead of doing combinations (like I mentioned in my post, i.e of AvsB, BvsC, and CvsA) and then taking overlapping features, this way is rather convenient. Essentially, I just wanna get features specific to type A compared to BandC, etc.

1
Entering edit mode

Exactly, in this case when you construct your matrix of contrast you should specify:

makeContrasts("myContrast" =  "A - (B + C)/2",
levels = dge\$design)


to perform such comparison