Hi all, I have been exploring different normalization options for a Medip- chip dataset we are analyzing, and I have a general question about the quantile normalization options in Limma. Our dataset shows a relatively pronounced batch effect, so the between-array normalization is a critical step. We have done within- array normalization using the Loess method first:, and then tested several quantile normalization options. The regular quantile normalization (done on the R and G channels) result in M-value distributions that still show relatively different quantiles, so we decided to normalize the M-values directly, passing the MA$M values: # Read data with Ringo RG = readNimblegen("targets.txt","spottypes.txt") # Normalize within arrays MA.Loess = normalizeWithinArrays(RG, method = "loess") # Normalize M-values MA.Loess.Mvalquantile = normalizeBetweenArrays(MA.Loess$M, method="quantile") Because this is not one of the built-in Limma functions, I was wondering whether there any obvious reason why this shouldn't be done. Any advice would be greatly appreciated. Gabriela Wlasiuk, PhD. Vercelli Lab Arizona Center for the Biology of Complex Diseases The University of Arizona