[BioC] questions of using Limma: should I include all the samples?

[Wu, Xiwei]

Dear all,

I am analyzing one set of data designed as follows:

Biological questions: what does an inhibitor (B) do to the hormone
stimulation effect (A) of a cell line.
Samples: 1) untreated (C)
               2) treated with Hormone alone (C+ A)
               3) treated with Inhibitor alone (C+ B)
               3) treated with hormone + inhibitor (C + A + B + AB)
There are three replicates each group (totally 12 CEl files).
Platform: Affymetrix GeneChip

I am assuming this is a 2X2 factorial desgn. My interpretation of the
biological question is the intersection of genes showing hormone and
inhibitor interaction effects (AB) and the genes responding to hormone (A)
alone. I am not trained as statistician, so correct me if I am wrong.

I am trying to use Limma with design matrix of 

                1 0 0 0
                1 0 0 0
                1 0 0 0
                0 1 0 0
                0 1 0 0
                0 1 0 0
                0 0 1 0
                0 0 1 0
                0 0 1 0
                0 0 0 1
                0 0 0 1
                0 0 0 1
                
to estimate the four coefficinet of C, C+ A, C+B and C+A+B+AB (of course, I
can estimate A, B, and AB directly using a different design matrix). 

Since the contrast of interest is A and AB, so the contrast matrix should
be:
-1  1  0  0 
-1 -1 -1  1

My question is: 
1) Are the design and contrast matrix correct?
2) I know this is a very naive question, but if I am only interested in
hormone only effect, can I just use the untreated and hormone alone treated
samples as the input (so instead of the 12 CEL files, only use the first 6
CEL files)? Will the analysis result be the same or different if not
counting the normalization-produced difference? If there is difference, is
that due to the difference of df?

Thanks in advance.

Xiwei Wu
Assistant Research Scientist
Beckman Research Institute
City of Hope National Medical Center
Duarte, CA 91010





-----------------------------------------------------------
SECURITY/CONFIDENTIALITY WARNING:  This message and any atta...{{dropped}}