Hello!

I recently did a biomark fluidigm high throughput qPCR experiment with a 96x96 format (i.e. 96 samples by 96 genes). My experimental design is rather complex with both control and experimental groups as well as timecourse data. I was hoping to use limma to analyze this data.

#Since my design is rather complex, I start by organizing my targets
targets<-readTargets("Targets.txt")

#Then I read in a table with my deltaCt values normalized to my housekeeing gene B2m
deltaCt<-read.table("B2m_limma.txt",header=T,sep="\t",row.names=1)

#Next I transform to log2 expression
y<- max(deltaCt) - deltaCt

#I set the different types of groups I would like to compare in different levels
flevels<-unique(targets$Group)
flevels
f<-factor(targets$Group,levels=flevels)

des<-model.matrix(~0+f)
colnames(des)<-flevels

fit<-lmFit(y,des)

contrast.matrix <- makeContrasts(
    Female7=FKO7-FWT7
    Male7= MKO7-MWT7
   Female15= FKO15-FWT15
    Male15= MKO15-MWT15
Female30= FKO30-FWT30
    Male30= MKO30-MWT30
,levels=des)

fit<-contrasts.fit(fit,contrast.matrix)
fit<-eBayes(fit)
options(digits=3)

My questions then become 1) Is this the correct way to read in this data and perform the analysis? 2) There are a few wells that did not work and/or there was no expression. How do I deal with these values in the data? (Their ct values are recorded as "999" as a default from the biomark software)

1. Yes, but I would also considering setting trend=TRUE in the eBayes step.
2. "did not work" and "no expression" are different things. If the former, then you set the value to NA. If the latter, then the Ct value (not the deltaCt) should be set to the maximum limit of the PCR system, usually about 40. If there really is a technical problem of some sort, then set the values to NA.