Comparing multiple groups in edgeR

Question

[sea.joson]

Hello!

I have 23 samples: treated vs control across time. I made(?) a design matrix as such.

   Treatment Time Group
1  None  0hr None.0hr
2  None  0hr None.0hr
3  None  0hr None.0hr
4  None  0hr None.0hr
5  None  3hr None.3hr
6  None  3hr None.3hr
7  None  3hr None.3hr
8  None  3hr None.3hr
9  None  6hr None.6hr
10 None  6hr None.6hr
11 None  6hr None.6hr
12 None  6hr None.6hr
13 None  9hr None.9hr
14 H2O2  3hr H2O2.3hr
15 H2O2  3hr H2O2.3hr
16 H2O2  3hr H2O2.3hr
17 H2O2  3hr H2O2.3hr
18 H2O2  6hr H2O2.6hr
19 H2O2  6hr H2O2.6hr
20 H2O2  6hr H2O2.6hr
21 H2O2  6hr H2O2.6hr
22 H2O2  9hr H2O2.9hr
23 H2O2  9hr H2O2.9hr

How do I compare multiple groups in edgeR? I understand that a design matrix would help. However, I don't seem to understand how to actually integrate it into the pipeline. I've been reading the user guide (3.3 Experiments with all combinations of multiple factors). I am quite new to bioinformatics and I have been trying to solve this issue on my own first. I've also tried

group <- c("1","1","1","1","2","2","2","2","3","3","3","3","4","5","5","5","5","6","6","6","6","7","7")

but then I only come up with

$comparison
[1] "1" "2"

Any help/guidance would be highly appreciated!

[Matthias Zepper]

You will need to use the GLM functionality of edgeR.

I found the user manual of the package is quite helpful in this regard, but since you mentioned that you already read 3.2 and the subsequent sections, maybe the examples in chapter 4 will help? In particular 4.2 corresponds exactly to your experimental design.

[sea.joson]

Hello! I have been working my way with the GLM approach and I think I finally understand it (at least more than before). I do have some questions though:

1. I made a design matrix and I did several comparisons with makeContrasts. Was this the right way?

1. Does this approach consider the number of replicates? Because for example, most groups have 4 samples but there is one group with only 2 samples, and another with only 1 sample.

[bkleiboeker]

1. Using makeContrasts to make relevant comparisons is a good and intuitive way to do what you are describing. I assume you went on to pass them into glmQLFTest?

2. Yes, assuming your design matrix was correct.

[sea.joson]

Yes I did use qlmQLFTest!

Thank you so much. I'm very happy for being able to have done this!