How to create "design" for RNA-seq experiment using limma?

Question

[Parham]

Hi!

I am using limma to analyse RNA-seq data. I have a problem I cannot sort it out by myself. This might be a very beginner question. I cannot figure out how to create design for voom . Its in chapter 15 of limma manual. What I created is:

> dge
An object of class "DGEList"
$counts
            Rep1.NoB1 Rep1.B1 Rep2.NoB1 Rep2.B1 Rep3.NoB1 Rep3.B1
SPAC212.11          1       0         1       0         3       0
SPAC212.10          1       0         0       2         0       0
SPAC212.09c         0       0         0       0         1       0
SPNCRNA.70          2       1         4       1         3       1
SPAC212.08c       216     274       229     312       267     435
7010 more rows ...

$samples group lib.size norm.factors Rep1.NoB1 1 12610138 0.9719185 Rep1.B1 2 9903882 1.0414141 Rep2.NoB1 1 8344326 0.9787473 Rep2.B1 2 11509548 1.0291455 Rep3.NoB1 1 13078338 0.9643794 Rep3.B1 2 18742285 1.0170714 ```

The experiment design is NoB1 is control and B1 is the treatment and it is done in three replicates. So Rep1.NoB1, Rep2.NoB1 and Rep3.NoB1 are replicates of control and the other three are treatments.

Next step is:

v <- voom(dge,design,plot=TRUE)

That I can't make it work! My question is what do I need to make the design matrix of? According to voom documentation it says "design matrix with rows corresponding to samples and columns to coefficients to be estimated. Defaults to the unit vector meaning that samples are treated as replicates."

What is this "coefficients"? And how can I group replicates together?

Thanks a lot in advance, I appreciate your help!

[sneha108ss]

Hi,

I am also facing a similar problem. How did you create the design matrix??

Answer 1

[Gordon Smyth]

Chapters 8 and 9 of the limma User's Guide explain how to make a design matrix for various experimental designs. The case studies in Chapters 17 and 18 give more examples.