Dear ALL,

i have an expression dataset(called data.trusted.eset)(affymetrix, hgu133a) which i have preprocessed and filtered and which i used as the input for my limma paired analysis to find DE genes. My basic question is, that i want to use the same expressionSet an an input(exprs(data.trusted.eset)) as a basic argument to a gene network co-regulation analysis,----but instead of probesets as rownames i have to replace them with unique gene symbols. Thus, remove in an appropriate way remove NAs but also duplicates.

I tried:    

affyUniverse <- featureNames(data.trusted.eset)

symbols <- hgu133aSYMBOL[affyUniverse]

symbol.Universe <- unique(as.character(symbols))

My main problem is then how to reduce my expression set with these symbols. Or my approach is totally wrong ? 

Any suggestions or ideas would be grateful

> gns <- select(hgu133a.db, featureNames(data.trusted.eset), "SYMBOL")
Warning message:
In .generateExtraRows(tab, keys, jointype) :
  'select' resulted in 1:many mapping between keys and return rows
> gns <- gns[!duplicated(gns[,1]),]
> ind <- !duplicated(gns[,2]) & !is.na(gns[,2])
> gns2 <- gns[ind,]

> table(table(gns2[,2]))

    1
13018

> data.trusted.eset.unique <- data.trusted.eset[ind,]

> featureNames(data.trusted.eset.unique) <- as.character(gns2[,2])

Dear Mr James W.MacDonald,

thank you for your valuable answer and code to implement it to my analysis. Before reading your answer in my question, i tried to write some small code to remove duplicate probesets based on median absolut deviation and keep the probeset with the bigger "mad" between the other same duplicates,as i think(correct me if im wrong) the duplicated function keeps the first incidence of a probeset . So i would like to see my code below and comment about my rationale of removing duplicates based on this characteristic(also no hard feelings about any mistakes or wrong syntax for my code because any comments or corrections would be essential to my improvement in R skills and knowledge):

symbols <- unlist(mget(featureNames(data.trusted.eset), env=hgu133aSYMBOL))
top2 <- topTable(fit2, coef="conditionCancer", number=nrow(fit2), genelist=symbols, adjust.method="fdr", sort.by="none") # where top2$ID holds the gene symbols for the probesets and with the arguments number and sort.by i have all my probesets with the same row as the rownames of the data.trusted.eset

m <- apply(exprs(data.trusted.eset),1,mad)

top2$MAD <- m
top2 <- na.omit(subset(top2, select=c(ID, logFC, adj.P.Val,MAD)))

top3 <- top2[order(top2$MAD, decreasing=TRUE),]

top3 <- top3[!duplicated(top3$ID),]

filtered <- data.trusted.eset[rownames(top3),]

rownames(filtered) <- top3$ID