[BioC] process one color microarray

[James W. MacDonald]

Hi Paz,

On 9/14/2011 5:55 PM, Paz Tapia Ramirez wrote:
> Hello, I have a question. I'm Working with one-color microarrays . I worked in 4 conditions different and each condition I have 4 replicates. Now, my question is  when I load the files to Bioconductor, I load as follows:
>   my.filenames<- c ("Condic1_repl1.txt",
>                                   "Condic1_repl2.txt",
>                                   "Condic1_repl3.txt",
>                                  "Condic1_repl4.txt",
>                                   "Condic2_repl1.txt",
>                                   "Condic2_repl2.txt",
>                                  "Condic2_repl3.txt",
>                                   "Condic2_repl4.txt",
>                                   "Condic3_repl1.txt"
>                                  "Condic3_repl2.txt",
>                                   "Condic3_repl3.txt",
>                                   "Condic3_repl4.txt",
>                                   "Condic3_repl1.txt",
>                                  "Condic3_repl2.txt",
>                                   "Condic3_repl3.txt",
>                                   "Condic3_repl4.txt")
>
> Subsequently, I realize the normalization procedure and some statistical calculations:
>   one.col1<-list (R = "gMeanSignal" G = "gProcessedSignal"
>              Rb = "gBGMedianSignal", Gb = "gProcessedBackground")
>   RG1<- read.maimages (my.filenames, source = "agilent", columns = one.col1, dec =".")
>    RG1<- backgroundCorrect (RG1, method = "half", offset = 50)
>    MA1<- normalizeBetweenArrays (RG1, method = "quantile")
>   fit1<- lmFit (MA1, design = NULL)

The design matrix indicates to lmFit() what is control and what is 
treatment. When you specify a NULL design matrix, lmFit() will just use 
a vector of 1s, which would be fine if you had two-color chips and no 
dye-swaps. Otherwise, you are just testing the hypothesis that the 
average expression of all samples is not equal to zero (which obviously 
isn't correct).

So if you have four conditions with four replicates (and I am assuming 
here that they are Biological replicates, not just different aliquots of 
the same sample), you want a design matrix with four columns. The 
simplest such design matrix (to me, anyway), computes the mean 
expression for each group, and then you can just make the comparisons 
you want.

cond <- factor(rep(1:4, each = 4))
design <- model.matrix(~ 0 + cond)
colnames(design) <- c("trt1","trt2","trt3","trt4")

contrast <- makeContrasts(trt2-trt1, trt3-trt1, trt4-trt1, trt3-trt2, 
trt4-trt2, trt4-trt3, levels = design)
fit <- lmFit(Ma1, design)
fit1 <- contrasts.fit(fit, contrast)
fit1 <- eBayes(fit1)

which will make all possible comparisons.

Alternatively, if you just want to compare all treatments to a control 
(and assuming your control is trt1).

design <- model.matrix(~cond)
fit <- lmFit(Ma1, design)
fit1 <- eBayes(fit)

In this case, all coefficients in the model will be e.g., trt2-trt1, 
trt3-trt1, trt4-trt1, so you don't have to specify contrasts directly.

Best,

Jim



>   fit1<- eBay (fit1)
>
> But my question is: how I can specify to bioconductor which files correspond to Control, or which correspond to microarrays with  treatment?
>
> Regards, Paz
>   		 	   		
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> Bioconductor mailing list
> Bioconductor at r-project.org
> https://stat.ethz.ch/mailman/listinfo/bioconductor
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor

-- 
James W. MacDonald, M.S.
Biostatistician
Douglas Lab
University of Michigan
Department of Human Genetics
5912 Buhl
1241 E. Catherine St.
Ann Arbor MI 48109-5618
734-615-7826

**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues