However, there are notable deviations from these genome-wide trends. First, there are a set of transcripts that show little variation between the 4 RNA samples and, in fact, these correspond to spike-in synthetic RNA added to each sample at a standardized concentration to act as controls having equal abundance in all experiments [40]. As expected, spike-in transcripts show significantly less variation across the 4 RNA samples than human genes (Fig. 3 b). Second, although technical effects are low for most genes, a small number of genes show high variation across laboratories and library constructions. In fact, the fraction of variation across laboratories correlates with the GC content of each gene (Fig. 3 c), and recapitulates the known role of GC content with reproducibility of RNA-seq data [8, 41–43].
