controls. A comprehensive evaluation of normalization techniques for RNA-Seq data has previously suggested the RPKM approach is ineffective and should cease to be used for evaluating differential expression (Dillies et al., 2013). Additionally, RPKM data may fail to adequately account for RNA composition bias (Robinson & Oshlack, 2010) or gene length (Bullard, Purdom, Hansen, & Dudoit, 2010) in the detection of differentially expressed features. Practical recommendations are available for generating fairly robust datasets (Dillies et al., 2013) and will continue to evolve as RNA-Seq is adopted in a larger number of laboratories. Selecting the appropriate statistical method for minimizing the effects of technical error will also depend upon additional known sources of systematic variation.
