450K complex design

Question

[Nicolas Rosewick]

Hi,

I've 48 samples that were analyzed using 450K (Infinium HumanMethylation450) for methylation and RNA-Seq (gene expression). Within these 48 samples I've multiple groups that can be summarized as:

Sample   Phase    Mutated
A        phase1   Yes
B        phase2   Yes
C        phase1   No
D        phase3   No
E        phase3   Yes

So for the phase variable three possibilities (phase1, phase2, phase3) and for the Mutated variable two possibilities (Yes, No). I want to analyse the data by:

1. Focus on the Phase variable (what are the differences between the different phases)
2. Focus on the Mutated variable (what are the differences between mutated and non-mutated samples)
3. Perform a complete analysis that takes into account both variable ( something like = ~Phase+Mutated)
4. Correlated the results with RNA-Seq data

I check a little bit in the literature and didn't find a package that allowed to perform a multi-group analysis (Point 1) and complex design (point 3). Anyone can advice me some interesting R packages that allow such analysis?

Thanks

Answer 1

[andrew.j.skelton73]

I'd suggest that you use the Minfi package for preprocessing, QC and normalisation. There's a function in there to look for differential methylation, but it's only for simple A vs B grouped comparisons. The good news is that it's just a wrapper function for limma, which can handle the complex design you've described. Remember to use M values to feed into Limma, not Beta Values!

[Nicolas Rosewick]

Ok great I'll give a try to limma. Thanks.

> edit : I used succesfully limma ;) thanks again