[BioC] Limma: design and contrast matrices

[Per Gregersen]

I have a 2 x 3 factorial cDNA microarray experiment which I would like to analyse in Limma. I need, however, some advise on the constuction of design and contrast matrices - which is not straightforward to understand for a non-statistician, as I am.
 
The experiment comprise two wheat lines (A,B - transgenic and wild type) with RNA samples harvested at 3 different time points (1,2,3 - develomental stages). The hybridisation experiment has been conducted as a circular loop: A1-->A2-->A3-->B3-->B2-->B1-->(and back to A1), and in addition there is a dye swap between A2 and B2, to balance the dyes. In total 8 arrays. 
 
I am primarily interested in the main effect of the transgene and any interaction effects between line and time. Secondarily, the main effect of time is also of interest, i.e. which genes go up or down between time1 and time2, and between time2 and time3 for both lines.
 
I hope, I can have some suggestions about the construction of design and constrast matrices, in order to analyse these effects in the best way.
 
Thanks.
 
Per Gregersen
Department of Plant Biology
Danish Institute of Agricultural Sciences
DK-4200 Slagelse, Denmark