Review

This paper is misguided. Despite the fact that it comes from a highly respected institute, the techniques it presents are archaic at best (division by the median was done as one of the first normalization techniques in existence) and it does not live up to real life scrutiny.

The paper of course is based around the R language and spends too much time on discussing import and export functionality of the package they developed. Too little attention goes towards the rationale of their normalization system. The paper fails to present a clear overview of how measurement errors are modeled and instead resorts to some magic wand waving which probably confused the reviewers of this paper.

As a practical experiment we tried to use this method to obtain useful data from 4 replicate screens and failed miserably. Instead modern techniques such as quantile normalization and an intensity dependent error model outclasses the presented method substantially.