preprocess.rle_normalise

preprocess.rle_normalise(X)

Median-of-ratios size factors (DESeq’s RLE).

Each sample is divided by the median, over genes, of its ratio to the geometric mean across samples. Genes with a zero anywhere drop out of the size-factor calculation – their geometric mean is zero and the ratio is undefined – which is the standard treatment and the reason RLE is robust to a few very highly expressed genes that would dominate a total-count scaling.