Dear All,

i have used the limma software package to implement an  paired statistical analysis for a dataset regarding DE expression between control & cancer samples. Heres my code:

library(limma)

library(hgu133a.db)

conditions <- data.trusted.eset$condition
condition <- factor(conditions, levels(condition)[c(2,1)])
pairs <- factor(rep(1:13, each = 2))

design <- model.matrix(~condition+pairs)
fit <- lmFit(data.trusted.eset, design)

fit2 <- eBayes(fit)

library(hgu133a.db)

symbols <- unlist(mget(featureNames(data.trusted.eset), env=hgu133aSYMBOL))

top <- topTable(fit2, coef="conditionCancer", number=nrow(fit2), adjust.method="fdr", genelist=symbols)

select <- top[which(abs(top$logFC) >1 & top$ adj.P.Val < 0.05),]

My main question(as a beginner in R/bioconductor), is because my platform is Affymetrix and some probesets match to the same gene, how i can remove these dublicates(select$ID column) which have the same gene 2 or more times and remove them, so that i can extract my DE list without gene dublicate symbols for further analysis ?

Thank you in advance

First of all, you probably shouldn't use "select" as a variable name, because it is already a function, and also the word is a verb, not a noun. I would suggest calling it "selected" instead.

Second, the function you're looking for is duplicated. You would do something like:

# Make sure the table is ordered with most the most 
# significant results at the top
selected <- selected[order(select$adj.P.Val),]
# Keep only the first occurrence of each ID
selected <- selected[!duplicated(select$ID),]

Thirdly, using separate cutoffs for fold change and significance (p-value or FDR) is not recommended. Consider using the treat function instead to test for significance relative to a nonzero threshold.

Thank you for your valuable answer and corrections !! . On the other hand, for the separate cutoffs you mean that i should not use simultaneously logFC and p-value cutoffs for sorting the results from topTable ? I thought that using logFC along with p-value cutoff gives more biologically meaningful results regarding the DE genes, instead of using just the adjusted p-value below a threshold. Moreover, one similar suggestion is mentioned in the paper of  JJ Chen et al., 2007{....."The fold-change criterion can always be used as a secondary criterion to facilitate the interpretation of biological significance. Some researchers may impose that the P-value and the foldchange are equally important; in such cases all genes satisfying both criteria are selected...."}.