[BioC] Limma: doing multiple paired t-tests in one go...

[Groot, Philip de]

Hello All,
 
I encountered a problem that I cannot easily solve, most probably because my knowledge of linear models is too restricted. The problem is that I want to do a paired t-test using limma, but that I want to fit multiple comparisons (using different patients!) simultanuously. The reason for this is that all .CEL-files in my experiment are fitted and this fit is used for the eBayes() command to maximize the advantage of using the eBayes approach.
 
I found in the bioconductor mailing list a somewhat related topic:
 
https://stat.ethz.ch/pipermail/bioconductor/2007-February/016123.html <https://stat.ethz.ch/pipermail/bioconductor/2007-February/016123.html> 
 
However, my problem is different. Instead of having multiple treatments over the same patients, I have multiple treatments over multiple patients (but still can do a paired t-test because before and after a single treatment is done on the same person).
 
For simplicity, let's assume that I have 2 diets and 2 patients for each diet. My pData(x.norm) looks like this:
 
                          sample       replicates sibship
A96_hA_09_113_1_base.CEL       1   Control_Diet_1     113
A96_hA_40_113_1_final.CEL      2 Treatment_Diet_1     113
A96_hA_10_114_1_base.CEL       3   Control_Diet_1     114
A96_hA_41_114_1_final.CEL      4 Treatment_Diet_1     114
A96_hA_01_101_2_base.CEL       5   Control_Diet_2     101
A96_hA_32_101_2_final.CEL      6 Treatment_Diet_2     101
A96_hA_02_103_2_base.CEL       7   Control_Diet_2     103
A96_hA_33_103_2_final.CEL      8 Treatment_Diet_2     103

 My design matrix (for a paired t-test) is calculated as follows (from the Limma user guide):
    Replicates <- factor(pData(x.norm)$replicates)
    SibShip <- factor(pData(x.norm)$sibship)
    design <- model.matrix(~SibShip+Replicates)

And the design matrix looks like this:
  (Intercept) SibShip103 SibShip113 SibShip114 ReplicatesControl_Diet_2
1           1          0          1          0                        0
2           1          0          1          0                        0
3           1          0          0          1                        0
4           1          0          0          1                        0
5           1          0          0          0                        1
6           1          0          0          0                        0
7           1          1          0          0                        1
8           1          1          0          0                        0
  ReplicatesTreatment_Diet_1 ReplicatesTreatment_Diet_2
1                          0                          0
2                          1                          0
3                          0                          0
4                          1                          0
5                          0                          0
6                          0                          1
7                          0                          0
8                          0                          1
attr(,"assign")
[1] 0 1 1 1 2 2 2
attr(,"contrasts")
attr(,"contrasts")$SibShip
[1] "contr.treatment"
attr(,"contrasts")$Replicates
[1] "contr.treatment"
 
 
As you can image, the comparisons I am interested in are Control_Diet_1-Treatment_Diet_1 and Control_Diet_2-Treatment_Diet_2. I might also be interested in Control_Diet_1-Control_Diet_2 and Treatment_Diet_1-Treatment_Diet_2, and so forth. My problem is that the current design matrix is rather complicated and that multiple interaction effects are somehow included, i.e. I cannot get individual effects by simply subtracting two factors in the design matrix (as I understand it). My question is: how can I create a contrast matrix that gives me the comparisons I am interested in? I am really looking forward to an answer!
 
Kind Regards,
 
Dr. Philip de Groot
Wageningen University