[BioC] limma design, 3 biological reps (two with a tech rep, one without)

Ryan Basom rbasom at fhcrc.org
Mon May 11 23:26:31 CEST 2009


Hi,

I'm trying to use limma to analyze the following:  Originally, there 
were 6 arrays (3 biological reps, each with a dye-swap hybridization).  
There was a technical issue with one of the arrays, leaving me with an 
'unbalanced' design:

SlideNumber    Name    FileName    Cy3    Cy5
1    PI-1_vs_DMSO-1    PI-1_vs_DMSO-1.txt    PI    DMSO

2    DMSO-2_vs_PI-2    DMSO-2_vs_PI-2.txt    DMSO    PI
3    PI-2_vs_DMSO-2    PI-2_vs_DMSO-2.txt    PI    DMSO

4    DMSO-3_vs_PI-3    DMSO-3_vs_PI-3.txt    DMSO    PI
5    PI-3_vs_DMSO-3    PI-3_vs_DMSO-3.txt    PI    DMSO



I'd like to attempt to include the sample without a technical replicate 
and would like to know if this this would be an acceptable way to handle 
this situation? ('pool' is a MAList containing 5 the arrays listed in 
the above order).


design <- cbind(PI1vsDMSO1 = c(-1, 0,0, 0,0), PI2vsDMSO2 = c(0, 1,-1, 
0,0), PI3vsDMSO3 = c(0, 0,0, 1,-1))
cont.matrix<-makeContrasts(PIvsDMSO = (PI1vsDMSO1 + PI2vsDMSO2 + 
PI3vsDMSO3)/3, levels=design)
fit<-lmFit(pool, design)
fit<-contrasts.fit(fit, cont.matrix)
fit<-eBayes(fit)


Thanks,

Ryan

-- 
Ryan Basom
Genomics Resource
Fred Hutchinson Cancer Research Center
Seattle, WA



More information about the Bioconductor mailing list