R Limma Package, Ebayes, Contrast Matrix

Question

[Cirrus]

Hello, I am using limma to analyse differential gene expressions. For modeling u need a design and contrast matrix. I just want to know whether anyone has experience with it.

Suppose that expressions are from wild type (WT) and mutants(M) and these are either stimulated (S) or unstimulated (U). For wild type I have 40 expressions values and for mutant 20. Should I correct for the different group size?

So when I want to know which genes respond differently in mutant compared to wild type, Which formula should I use for the contrast Matrix:

Diff=(M.S-M.U)-(WT.S-WT.U)

or

Diff=(M.S/20-M.U/20)-(WT.S/40-WT.U/40)

[Martin Morgan]

The best place to ask this is the Bioconductor mailing list, after reading the excellent limma user guide.

Answer 1

[seidel]

Use the first formula. Limma accounts for different sample sizes. It's common to throw out an array here or there as an outlier, so ending up with unbalanced sample sets is not uncommon. Though just as an exploration you might reduce the wt set to match, and see how dramatic the change is (the results should be similar, you just have less replication).

Answer 2

[Martin Morgan]

No need to include sample size. The best place to ask these questions is still the Bioconductor mailing list