How to further process inSilicoMerging data in R?

Question

[blubb1]

I'm working on a microarray analysis using R/bioconductor (oligo package). I need to merge two datasets (affymetrix pd.hta.2.0.st and pd.hugene.2.0.st) and did this using inSilicoMerging. However, if I now look at the expressionSet, basically all information of the probes/genes is gone, so how am I supposed to further analyse the combined datasets?

the data looks like this:

01.CEL  02.CEL  03.CEL

1 549.326.420.227.016 485.001.635.030.679 383.486.118.570.383

so just insane high numbers (still intensity values or what exactly are these number representing?) and instead of a probe identifier just for example a 1.

thanks for any help! :)

Answer 1

[andrew.j.skelton73]

If I were you, I'd normalise each dataset independently, and make sure you have two expression sets with common identifiers. I'd recommend you use nuIDs, then you're working with the same capture sequences at least. inSilicoMerging isn't magic, you need to help it out, and it's likely that because you've given it different probe identifiers for each platform, it's just labelled them numerically.

There are caveats to this kind of methodology, particularly if your end goal is to "increase your experimental power" for a differential expression analysis. Firstly, you need at the very least some samples to be of the same phenotype across both platforms, this allows you to estimate the variance of your technical effect (in this case platform). Secondly, your method of "merging" or technical correction is an important judgement to make, I'd personally suggest you use inSilicoMerging's "none" option, and add a term in your linear model to account for the platform effect.