My goal is to perform biomarker discovery on 123 MIAME-compliant samples from Invitrogen's ProtoArray v5.0.
So far, I have used the PAA Bioconductor package to upload the raw data files and perform background correction, batch adjustment, and normalization.
Before applying multivariate feature selection, I would like to eliminate features (antibodies) whose expressions are unreliable or represent noise in order to 1) lessen the impact of the curse of dimensionality and 2) decrease run times for feature selection in R to something more managable.
My brief experience with this task is limited to filtering Affymetrix gene expression data which came with detection calls. For one project, I filtered so that the only probe sets that remained had both a minimum of 25% Present calls in at least 1 out of the 2 class AND had a range of expression values greater-than-or-equal-to the trimmed mean.
How would I go about this task in the context of ProtoArray data?
