This can be achieved either by doing an additional normalization step after scaling, but ideally by using a between-microarray normalization approach from the beginning, such as applying T-quantile normalization as demonstrated here.
We conduct microarray normalization using a lowess normalization approach for cDNA data and a robust normalization approach for Affymetrix data [ 24].
Because microarray data use a different normalization approach based on a larger part of the transcriptome, we conclude that omitting endogenous control genes will give measurements more in accordance with actual concentrations.
In addition, microarrays should be normalized both within and between the technologies, incorporating a normalization approach across all of the data in the experiment [ Kuo et al., 2006].
These qualitative discrepancies between both approaches uncover a fundamental problem of microarray normalization with no satisfactory solution yet (see, e.g., [ 45]).
The approach should thus be used with caution and as a complement to classical microarray normalization methods that do not involve correction for histone density, which are also used in this study.
Analysis approaches have been developed specifically for RNA-Seq [ 17- 21], yet repurposed microarray normalization, statistical analysis, and DEG identification methods outperform several RNA-Seq specific analysis tools [ 20, 21].
Many tools are now freely available to aid investigators with microarray normalization and selection of internal reference genes to be used for independent corroboration of microarray results.
We sought an alternative normalization approach, based on the idea of choosing a minimal rank difference or rank-invariant miRNA set from within the data, a strategy commonly employed in microarray analysis [55].
A detailed normalization approach is presented in Text S1.
