I am analyzing microarray data using the Bioconductor limma package in R. The experiment is a direct two-color experiment comparing two genotypes. There are two probes for each gene and I would like to analyze the ratio of the probe signals. I am using the standard workflow of first fitting a linear model and then applying the Bayesian hierarchical modeling:
fit <- lmFit(normalizedValues, designMatrix)
fit <- eBayes(fit)
So far, I have calculated the ratios of the two probes from the raw data for each replicate and then fed it into the analysis pipeline. However, this does not reflect the variance associated with the raw values that were used to calculate the ratio. My guess is that I am underestimating the variance. Is there a way to accurately perform such an analysis using the limma package? I can calculate an estimate of the variance using Fieller's theorem but I don't know how to feed that into the pipeline without seriously tinkering with the limma functions.
Thanks for your help!