**70**wrote:

I am fairly new to microarray analysis. I have 40 arrays from four phenotypes and would like to compare the gene expression among the four subtypes.

For this i used bead array (since the arrays are illumina based), preprocessed with BASH/HULK and summarised it and ran neqc for background correction and normalisation and finally have log2 expression data as an ExpressionSetillumina.

I then use this for carrying out differential expression analysis using Limma. I would like to compare the gene expression of one subtype with the others. For this I first use the design matrix and define the arrays corresponding to a subtype with one and the rest 0. I then compare the groups. Note that there are different number of repilcates in each subgroup.

Sadly I donot get any differential expression which is very unlikely and I am sure there is something wrong.

Any comments?

To see if anything is wrong, you will have to post the code used. From your description, it is not obvious just how you made your design matrix, and which coefficient you are checking against.

1.4kHi, I agree... please see the comments I have posted it there.

70Here is what I have done: Phenos is the phenotypes whose RNA is measured and Eset is the expression set. PLease dp let me know if any relevant information is missing.. Here I compare the Pheotypes in pairs.. I also tried comparing one Phenotype to the other three. That also did not work. I also tried different normalizations apart from neqc.

70I am not familiar with Illumina beadchip arrays, but have you checked if any of the arrays are outliers or weird, and how the distributions look after normalisation?

Also, the results from topTable looks very odd. By default, it should only check coef=1, i.e. Group1 vs Group2. Maybe you should also try asking on the Bioconductor mailing list.

1.4kYa I checked for outliers... I think I will ask the mailing group as you have suggested.

70