[BioC] gcRMA on a GeneFeatureSet from Affy MoGene 1.0 ST v1?

[Jean Hausser]

Hi all,

I'm trying to analyze Affy Mouse Gene 1.0 ST v1 chips with  
BioConductor 2.2 and R 2.7.1. I had a pipeline to process the previous  
affy chips (MOE430, HGU133 Plus 2.0, etc.) using the affy/gcrma/limma  
trio but this pipeline won't eat the newer Affy Mouse Gene 1.0 ST chips.

Thanks to James McDonald's answer to Hyeong-Min on the BioC  
mailing-list two weeks ago, I could use the pdInfo package to read in  
the CEL files into R using oligo. But I'm struggling to run the raw  
data through gcrma. oligo implements RMA but I'd really like to use  
gcRMA to compute corrected probe-level intensities. More specifically,  
I'd like gcrma to use the "antagenomic" probes from the MoGene chip  
when estimating non-specific binding.

I first tried the following:


library(gcrma)
library(oligo)
library(pd.mogene.1.0.st.v1)
outputDir <- file.path(getwd(), "chipDBI");

gfs <- read.celfiles(expDesign$filename);
chipAnnot <- read.table("MoGene-1_0-st-v1.probe.tab", h=T, sep="\t");
bkgProbes <- chipAnnot$Probe.ID[as.character(chipAnnot$category)
                                 == "control->bgp->antigenomic"];
eset <- gcrma(gfs, normalize=F, affinity.source="local", NCprobes=bkgProbes);


where "MoGene-1_0-st-v1.probe.tab" is the probe annotation table from  
Affymetrix.
The problem is that gcrma expects an "AffyBatch" affy object while I'm  
providing a "GeneFeatureSet" oligo object.

Therefore, I started hacking a gcrma() function that would work with a  
GeneFeatureSet. But I quickly ran into trouble because gcrma has a lot  
of sub-functions, many of them calling to pm(), mm() and IndexProbe(),  
which are AffyBatch methods that are not implemented in  
GeneFeatureSet. And even in the first step bg.adjust.optical(), I  
can't figure how to overwrite probe intensities in a GeneFeatureSet.  
For instance, "exprs(gfs)[1,1] <- 1" says there's no "exprs<-" method  
for GeneFeatureSets but still, exprs(gfs) does return the probes  
intensities...

I'm starting to think this is either a bone that's too hard for me to  
eat in a reasonable amount of time (I'm not really that familiar with  
the BioC classes and object oriented programming in R) or there is a  
much easier way I don't know about? So does someone maybe already have  
a version of gcrma that's working with GeneFeatureSet? Or, is it a  
feature that is planned for the next version of BioC?


Thanks in advance!

Cheers,


-- 
  Jean Hausser                                +41 61 267 15 73
  Zavolan Group, Bioinformatics, Biozentrum, Universität Basel