I am doing differential expression on a fairly large RNA-Seq dataset, where multiple organs are obtained from the same mice. I would like to model the effects of individual mice using limma's random effects via
duplicateCorrelation. As library sizes are very consistent across all samples (and I might want to play around with
genas), I was thinking of using limma-trend.
I've seen some posts on using
voomWithArrayWeights together with
duplicateCorrelation, but nothing on limma-trend.
Given that a standard limma-trend pipeline goes as follows (starting from a count matrix
EM and design
dge <- DGEList(EM) dge <- calcNormFactors(dge) v <- cpm(dge, log=TRUE, prior.count=3) fit <- lmFit(v, design=mod) eb <- eBayes(fit, trend=TRUE, robust=TRUE) decideTests(eb) topTable(eb) ...
How would this be modified to accommodate