Using Limma to normalize data sets from microarray studies

Question

[Natasha]

I'm using Limma to normalize Affy data sets from 2 experimental studies performed using microarray ,

To check whether the steps that I follow is correct, I am checking whether the box plot that I obtain after processing the RAW file is the same as the boxplot obtained from GSE. GSE:

library(Biobase)
library(GEOquery)
eset <- getGEO('GSE20966')[[1]] 
boxplot(exprs(eset), outline=FALSE)

RAW file:

library(gcrma)
library(limma)
downloadedAffyFiles <- list.files(path = "../Data/GSE20966_RAW/", pattern = "CEL.gz$",full.names=TRUE)
AffyData <- ReadAffy(filenames = downloadedAffyFiles)
eset <- gcrma(AffyData)
Data <- exprs(eset)
boxplot(exprs(eset), outline=FALSE)

Since both the plots are different, I understand the difference comes out of the normalization method that is used. Could someone suggest what changes have to be made in the above syntax? How can I set the method of normalization?

[WouterDeCoster]

How to add images to a Biostars post

I assume this is microarray data, but you don't mention this. It doesn't hurt to be specific.

Please use tags appropriately, as such experts can easily find your question and help you. In this case microarray and limma would make sense.

[Natasha]

Hope I made all the changes

[ATpoint]

You have to pass the full link of the picture to the image buttom (right next to the code (10101) field). I made the changes for you this time.

Answer 1

[Mark]

You should read the manual. It is extremely detailed and helpful. See section 6.2: https://www.bioconductor.org/packages/release/bioc/vignettes/limma/inst/doc/usersguide.pdf

When you read the data in using gcrma function (from the gcrma package) you are using quantile normalization method:

https://www.rdocumentation.org/packages/gcrma/versions/2.44.0/topics/gcrma

Set normalize = FALSE and then using the limma package you can normalise with whatever method is most suitable:

 MA <- normalizeWithinArrays(RG)
 MA <- normalizeWithinArrays(RG, method="robustspline")

Again, I highly suggest you read the manual it is extremely helpful and informative.

[Natasha]

Hi, I had a chance to look at section 6.2 before. The section heading mentions it's for two color data. The data set that I have is one color data. Also, I couldn't understand how the target file that is used to create RG in the syntax mentioned by you has to be generated. In my case, rownames(exprs(eset)) directly gives the probeset id's. Should this go into the targets.txt file?

I'm really sorry for the naive questions. I tried my best to understand from the manual before posting my question. This field is new to me, I'm facing challenge in understanding things at the first go. I'm referring to section 8.2, which explains about creating a designmatrix.

> design <- model.matrix(~ 0+factor(c(1,1,1,2,2,3,3,3)))

I am stuck at this step.If I understand correctly, 1,2,3 refers to state like disease or control. I couldn't parse this information for GSE20966 from the eset created using eset <- gcrma(AffyData,normalize=TRUE)