[BioC] [maNorm] Normalization a complex experiment...

Marcelo Luiz de Laia mlaia at fcav.unesp.br
Wed Jan 21 22:26:41 MET 2004

In the Tue, 20 Jan 2004 10:42:40 +1100
Gordon Smyth <smyth at wehi.edu.au> write:

GS> Heatdiagram may help you visualize your results, but what you really need 
GS> is the F-statistic computed by the classifyTests() function. This is not 
GS> yet explained in the User's Guide. Can you consult a local statistician for 
GS> help who knows a little about linear models and contrasts?
GS> Gordon

Hi Gordon and All,

I find in the html_help and in the user's guide and I accomplished
the following test with my data.

If is possible, I would like you to verify if it is correct, because
 I am not sure of that.

The experiment design is:

           1day 2day 3day

           Rep1 Rep1 Rep1
Un Treated Rep2 Rep2 Rep2
           Rep3 Rep3 Rep3

           Rep1 Rep1 Rep1
   Treated Rep2 Rep2 Rep2
           Rep3 Rep3 Rep3

It follows the script:

> library(limma)
> RG <- read.maimages(files, columns=list(Rf="DataVol",Gf="CtrlVol",
+ Rb="DataBkgd",Gb="CtrlBkgd"))
> show(RG)
> summary(RG$R)
> genes.names[1:10,]
> printer <- list(ngrid.r=4, ngrid.c=5, nspot.r=16, nspot.c=24, ndups=2,
+ spacing=1, npins=20, start="topleft")
> printer
> MA <- normalizeWithinArrays(RG, method="none", printer)
> boxplot(MA$M~col(MA$M))
> MA <- normalizeWithinArrays(RG, printer)
> boxplot(MA$M~col(MA$M))
> MA.fa <- normalizeBetweenArrays(MA,method="scale")
> boxplot(MA.fa$M~col(MA.fa$M))
> design <- model.matrix(~ -1+factor(c(1,1,1,2,2,2,3,3,3)))
> colnames(design) <- c("time1","time2","time3")
> fit <- lmFit(MA.fa,design)
> contrast.matrix <- makeContrasts(time2-time1, time3-time2,time3-time1,levels=design)
> fit2 <- contrasts.fit(fit,contrast.matrix)
> fit3 <- eBayes(fit2)
> time2.time1 <- topTable(fit3, coef=1, adjust="fdr")
> time3.time2 <- topTable(fit3, coef=2, adjust="fdr")
> time3.time1 <- topTable(fit3, coef=3, adjust="fdr")
> clas <- classifyTests(fit3)
> vennDiagram(clas)

If the script is correct, I obtained genes common to the time 2 and
time 1, time 3 and time 2 and time 3 and time 1. However, I didn't
obtain any gene common to the three times.

Is there as leaving the test a little less rigorous?

Maybe I find at least about 5 genes common to the three times! Or it is not
 a good ideia?

Thanks very much

Best wishes



More information about the Bioconductor mailing list