dear Heidi, dear all, I've just tried to normalize Taqman microfluidic card data with the HTqPCR package using the "geometric.mean" method and was puzzled by its results, since, instead of making the samples more comparable, they become more different. actually, I think I spotted the problem in the code: geometric.mean = { geo.mean <- apply(data, 2, function(x) { xx <- log2(subset(x, x < Ct.max)) 2^mean(xx) }) if (missing(geo.mean.ref)) geo.mean.ref <- 1 geo.scale <- geo.mean/geo.mean[geo.mean.ref] data.norm <- t(t(data) * geo.scale) if (verbose) { cat(c("Scaling Ct values\n\tUsing geometric mean within each sample\n")) cat(c("\tScaling factors:", format(geo.scale, digits = 3), "\n")) } so basically, the scaling factor is a ratio between the average Ct value of a sample and the average Ct in the reference sample. so if the Ct values are on average higher in the sample compared to the reference the scaling factor is > 1. For normalization, however, this scaling factor is multiplied to the Ct values of the sample, making the difference bigger. Thus, I suggest to change the data.norm <- t( t( data ) * geo.scale ) to data.norm <- t( t( data ) / geo.scale ) cheers, jo -- Johannes Rainer, PhD Applied Bioinformatics Group, Division Molecular Pathophysiology, Biocenter, Medical University Innsbruck, Innrain 80/82 II, 6020 Innsbruck, Austria and Tyrolean Cancer Research Institute Innrain 66, 6020 Innsbruck, Austria Tel.: +43 (0)512 9003 70961 Email: johannes.rainer at i-med.ac.at johannes.rainer at tcri.at URL: http://bioinfo.i-med.ac.at