A question occur when using DESeq2 on a complex design
1
0
Entering edit mode
@466115b6
Last seen 9 days ago
United States

Hi everyone here,

It seems have a problem when I trying to creat a new post here. So I put the question on another website: https://www.biostars.org/p/9599648/

Hoping for any information about this problem. Thank you very much!

DESeq2 rnaseqGene RNASeq • 517 views
0
Entering edit mode

It would be very appreciate if if you could check this question out. Michael Love

2
Entering edit mode
@mikelove
Last seen 1 day ago
United States

Note that:

(A/B) / (C/D) = AD / BC

with the right-hand side written in DESeq2's syntax as:

list(c("A","D"), c("B","C"))

I'm not sure you've got your contrast right.

0
Entering edit mode

Thanks very, very much for your answer.

I would like to start with an A1 vs A0 comparison, then a B1 vs B0 comparison, and finally realize ((A1 vs A0) vs (B1 vs B0)). I designed the comparisons by referring to the book 'RNA-seq Data Analysis: A Practical Approach', where it is mentioned that such complex comparisons can be realized using DESeq2. However it now looks like I don't fully understand how it actually works. In my experimental design, unlike in the book, my groups A and B have their own controls, A0 and B0. The book uses a common control for the complex comparison. So, is it possible that the comparison I want to make cannot be realized using DESeq2 when using different controls?? Or can you help me such as tweaking the code to achieve the comparison I want? Thanks very very much again!

Create the DESeqDataSet object using patient, run, time, and treatment as covariates: dds <- DESeqDataSetFromMatrix(countData = countData, colData = meta, design = ~patient + run + time + treatment)

If you had wanted an overall comparison of the 24 and 48 h time points (including all treatments and patients), you could have used results(dds,'time_48h_vs_24 h') instead, and so on. What if we wanted to a more speciffc or complex comparison? For example, we might want to compare the difference between DPN and control, on the one hand, and the difference between OHT and control, on the other hand, that is, a difference of differences which would tell us something about DPN treatment effects that are not seen in OHT treatment. To do this, we could encode the (DPN vs. control) vs (OHT vs. control) using the contrast argument of the results function (note that this is only available in DESeq2 versions 1.1.24 and above). As the vector returned by resultsNames() has 'treatment_DPN_vs_Control' as its seventh entry and 'treatment_OHT_vs_Control' as its eighth entry, we can do res <- results(dds,contrast=c(0,0,0,0,0,0,1,-1))

1
Entering edit mode

Just want to make sure you get this one point, you had:

list(c("A_1", "A_0"), c("B_1", "B_0"))

this is numerically:

(A1 x A0) / (B1 x B0) = (A1/B1) / (B0/A0)

Notice the ratios on the right-hand side.

For questions about how to optimally set up the statistical analysis and design choices, you should really work with a local statistician or someone familiar with linear modeling in R. I have to restrict my time on the support site to software issues and software-related questions, and can't respond to in depth questions about optimal analysis plans.

0
Entering edit mode

Thanks for the quick and clear answer, I think I understand the problem with it. Much, much appreciated!