ABSTRACT
DE analysis of RNA-seq data differs from microarray DE analysis in that the observed data are in the form of discrete counts generated from a sampling process, while microarray measurements are continuous measurements of a fluorescence signal. One aspect of this is that, because RNA-seq is a sampling procedure, there is a certain amount of “real estate” (the total number of all reads from the sequencing instrument) that the actual transcripts in the sequencing library have to “share.” This means that highly expressed transcripts will often make up a large amount of the sequencing library, and in a shallow sequencing experiment less expressed genes may not be represented in the final data even though they were present. By contrast, microarrays are not constrained to such a “zero-sum game,” although they of course have other limitations. An attractive feature of RNA-seq, though, is the possibility to re-sequence the same library to potentially recover more expressed transcripts.
