[BioC] duplicateCorrelation dye-swap technical replicates

[Roberta Sirovich]

Dear all,

I am working with limma package for the analysis of differential 
expression in agilend double color chips.
My data come from a common reference experiment and the design matrix is 
the following:
WT   MU
1   0
-1   0
0   1
0   -1
1   0
-1   0
0   1
0   -1
Since rows 1 and 2 are technical dye-swap replicates (as rows 3 and 4, 5 
and 6, 7 and 8), I am trying to use duplicateCorrelation with the block 
argument as follows:
rep<-(1,1,2,2,3,3,4,4). But there's something that I don't understand.
1. if I give the following design matrix (instead of the one written above)
(1,-1,1,-1,1,-1,1-1), the result in corfit$consensus changes (but not 
very much) why?
2. if I give as design matix the abs(.) of the matrix above, the result 
in confirt$consensus changes very much. why?
I thought that duplicateCorrelation needs the design matrix to be able 
to distinguish between WT and MU. I thought it was't necessary to say 
which chip is dye-swap (it should come from the logratio values?). And 
why the result changes if I give the vector in 1.?
What exactly duplicateCorrelation calculates?
Thak you very much!
Roberta Sirovich

-- 



-----------------------------
Dr. Roberta Sirovich
Department of Mathematics
University of Torino
Via Carlo Alberto 10
10123 Torino, Italy
tel: +39-011-6702850
fax: +39-011-6702878