Figure 4

Scaled quantile normalization: concept. If two data columns with different numbers of values, derived from two A1 and A2 samples, respectively, are individually sorted by magnitude of expression to obtain the mean value for all values with the same rank, i.e. in the same row (quantile normalization), the highest values in the sample A2 will be aggregate to the intermediate values in the sample A1. Proportional scaling of A2 ranks aligns them to A1 values located in analogous ordered positions with respect to each sample whole distribution (scaled quantile inter-sample normalization), allowing low, intermediate and high values to be aggregated with suited corresponding values from the other sample(s).