[BioC] 2fold sigchanges from limma

Ivan Baxter ibaxter at purdue.edu
Sat Dec 2 00:04:51 CET 2006


Greetings-

I am using limma to analyze a fairly simple multiple treatment 
experiment and I would like to pull out all the genes which are 
significantly expressed and change at least two fold. I have figured out 
a way to do this, but it is a little complicated and I am running into 
problems downstream when I try to format output tables using annaffy 
(described below).  Is there a simple way to filter the output of 
topTable to only get the genes which are significant and change more 
than a give fold-change cutoff?


The way that I am going about this is....

cont.matrix <- makeContrasts(
    EGvsC    = EG.C.C.C - C.C.C.C,
    ECvsC    = C.EC.C.C - C.C.C.C,
    EGECvsC  = EG.EC.C.C - C.C.C.C,
    GTvsC   = C.C.G.C - C.C.C.C,
    GT_APvsC   = C.C.G.A  - C.C.C.C,
    GT_APvsGT = C.C.G.A  - C.C.G.C,
    levels=design)
fit2 <- contrasts.fit(fit, cont.matrix)
fit2 <- eBayes(fit2)
results <- decideTests(fit2)
idx <- abs(fit2$coefficients) > 1
#make a dataframe like results where 1/-1 indicates sigchange greater 
than 2 fold
comb <- data.frame(gene = rownames(results))
for(i in 1:length(rownames(results))){
    for(j in 1:length(colnames(results))){
    
        if(results[i,j] != 0 & idx[i,j] == "TRUE"){
            comb[i,j+1] <- results[i,j]
           }
       else{comb[i,j+1] <-0}
    }
}

#but then I want to make an output table with gene names, gene 
annotations, fold change, pvalue and the expression values #across the 
arrays......


cax1_genes <- unlist(as.list(comb$gene[comb$GTAPvsC != 0]))
syms <- unlist(mget(cax1_genes, hgu133a2GENENAME))
test <- match(cax1_genes, geneNames(human.eps2))
anncols <- aaf.handler()[c(1:4,7)]
anntable <- aafTableAnn(geneNames(human.eps2)[test], "hgu133a2", anncols)

#now I need to get the fold- change and adj.p.value for each gene, and 
here is where I run into trouble.
#I tried pulling out all the significant changes using topTable and then 
pulling out the list of genes that met my criteria, but...
contp <- 5 # this is the contrast which is being tested
caxsig <- length(which(p.adjust(fit2$p.value[,contp], method = "BH") < 
0.05))
cax1sig <- topTable(fit2, coef =contp, number =caxsig, adjust.method = 
"none", sort.by = "p", resort.by = "M")
testtable <- aafTable("log2 change" = format(cax1sig$M[cax1sig$ID == 
cax1_genes], digits = 2),
                                    "pval" = 
format(cax1sig$adj.P.Val[cax1sig$ID == cax1_genes], digits = 2))
anntablep <- merge(anntable, testtable)


# doesn't work, I get the following error:
 > testtable <- aafTable("log2 change" = format(cax1sig$M[cax1sig$ID == 
cax1_genes], digits = 2), "pval" = format(cax1sig$adj.P.Val[cax1sig$ID 
== cax1_genes], digits = 2))
Warning messages:
1: longer object length
        is not a multiple of shorter object length in: cax1sig$ID == 
cax1_genes
2: longer object length
        is not a multiple of shorter object length in: cax1sig$ID == 
cax1_genes

If this is actually a good way to pull out the two fold genes, could 
anyone tell me what I am doing wrong with this last step?

thanks in advance.

Ivan



 > sessionInfo()
Version 2.3.0 (2006-04-24)
i386-pc-mingw32

attached base packages:
 [1] "grid"      "splines"   "tools"     "methods"   "stats"     
"graphics"  "grDevices" "utils"     "datasets"  "base"    

other attached packages:
     annaffy         KEGG           GO     hgu133a2 RColorBrewer  
geneplotter     annotate       hexbin   colorspace      lattice   
genefilter
     "1.4.0"     "1.12.0"     "1.12.0"     "1.12.0"      "0.2-3"     
"1.10.0"     "1.10.0"      "1.6.0"        "0.9"     "0.13-8"     "1.10.1"
    survival        limma        gcrma  matchprobes         affy       
affyio      Biobase
      "2.24"      "2.7.3"      "2.4.1"      "1.4.0"     "1.10.0"      
"1.0.0"     "1.10.0"



-- 
**************************************************************
Ivan Baxter
Research Scientist
Bindley Bioscience Center
Purdue University
765-543-7288
ibaxter at purdue.edu



More information about the Bioconductor mailing list