Why are normalization methods not interchangeable?

SARTools is a R package associated with two R script templates which allow one to perform differential analysis with either DESeq2 [1] or edgeR [2], the normalization method employed being the one associated with the package used. The purpose of this additional file is to show that normalization methods are not interchangeable between statistical models without adequate transformation. Despite this has been recalled in [3] some users of R packages for differential expression are still not aware of that. This is an important issue as DESeq(2) and edgeR use normalization factors in two different ways : DESeq(2) integrates the size factors in the calculation of the mean of the Negative Binomial distribution used to model raw counts when edgeR normalizes library sizes and includes them as an offset in the statistical model for differential testing. Therefore, normalization factors should be computed and used with regard to the statistical test that follows.