Hello,

I have a question about some proteomics data that I am analysing in R. I have log2 transformed the intensity data (from mass spec) and would like to perform median normalisation on this dataset. I came across the function "normalizeMedianValues(x)" in Limma. In my case my matrix x contains the log transformed intensities where each column is a different sample (in this case patients) and each row is a different protein.

Is this function okay to use in this situation? It seems like it was originally designed for normalising between different arrays.

Any advice is much appreciated!

Thanks, Sandra

You could, but you need to 'unlog' first, as normalizeBetweenArrays (which is the function you should be using, rather than a helper function like normalizeMedianValues). Assuming your data are log2 transformed, you could do

x <- 2^x
normx <- normalizeBetweenArrays(x, "scale")

And go on from there. Note that normx will now contain log2 transformed data.

You could try median polish of ratio, tunable by which samples you choose for central tendency. The TAMPOR package has been published by Johnson ECB, et al, Nat Med, (2020) and is available on GitHub, designed for proteomic data with global internal standard (GIS) samples, but it works even for biological replicates of a particular treatment or disease (or control) group present in each batch.