[BioC] Help to confirm Design matrix and Between array Normalization

[Gordon K Smyth]

Dear Veerendra,

Your design matrix looks fine as far as I can see.

Between array normalization would make no difference to your results. 
That is only required if you want to try separate channel analysis.

Best wishes
Gordon

> Date: Thu, 12 Aug 2010 19:10:10 +0530
> From: Veerendra GP <gpveerendra09 at gmail.com>
> To: bioconductor at stat.math.ethz.ch
> Cc: prashantha.hebbar at manipal.edu, shama.prasad at manipal.edu
> Subject: [BioC] Help to confirm Design matrix and Between array
> 	Normalization.
>
> Dear friends,
>
> I have DNA differently methylated data of (done with differentially
> methylated hybridization technique) three cancerous experiments. The
> experimental design is as follows:
>
> Normal vs Tumor: wherein Normal is labeled with cy3 and Tumor is labeled
> with cy5.
> Normal vs Pre malignant: wherein Normal is labeled with cy3 and Pre
> malignant is labeled with cy5.
> Tumor vs Pre maligmant: wherein Tumor is labeled with cy3 and pre malignent
> is labeled with cy5.
>
> I wanted to fit linear model for this experimental design. I have gone
> through the Limma documentation and done analysis. But, as design matrix
> need to be more perfect, I would like to confirm the matrix design with your
> opinion. I also would like to know fitting between array normalization in
> this context will be meaningfull or not? So, I request you to help me in
> this context.
>
> Here is my sessional info :
>
> library(limma)
> targets<- readTargets("/home/veerendra/MicroarrayData/target.txt")
> RG <- read.maimages(targets$SlideNumber, source="agilent",
> path="/home/veerendra/MicroarrayData/DATA")
> status <- rep("gene", nrow(RG$genes))
> status[grep("NC1_*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("NC2_*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("LACC*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("PC_*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("DarkCorner",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("HsCGHBrightCorner",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("NegativeControl",RG$genes[,"SystematicName"])] <- "cntrl"
> status[grep("SM_*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("DCP*",RG$genes[,"ProbeName"])] <- "cntrl"
> status[grep("unmapped",RG$genes[,"SystematicName"])] <- "cntrl"
> RGnc <- RG[status!="cntrl",]
> MA<-normalizeWithinArrays(RGnc,method="loess",bc.method="normexp")
> design <- modelMatrix(targets, ref="N")
> N P T
>     P  T
> [1,] 0  1
> [2,] 1  0
> [3,] 1 -1
>
> fit <- lmFit(MA, design)
> contrast.matrix <- cbind("N-T"=c(1,0),"N-P"=c(0,1),"P-T"=c(1,-1))
>
>  N-T N-P P-T
> [1,]   1   0   1
> [2,]   0   1  -1
>
> rownames(contrast.matrix) <- colnames(design)
> fit2 <- contrasts.fit(fit, contrast.matrix)
> fit2 <- eBayes(fit2)
> tb<-topTable(fit2, adjust="BH", n=200000)
> write.table(tb,file="testtb.txt", sep="\t")
>
> Thanking you in anticipation.
>
> Regards,
>
> Veerendra.

______________________________________________________________________
The information in this email is confidential and intend...{{dropped:4}}