As expected, variation across the 4 RNA samples is the major axis of variation, explaining a median of 87.5% of variation in expression (Fig. 3 a). But the real interest is in the sources of technical variability. The fact that the technical variables laboratory (2.93%), library (2.55%), flowcell (0.0057%), and lane (0.0000000038%) explain a small fraction of the total variation indicate that these RNA-seq experiments were highly reproducible genome-wide. Interpreting these values in terms of the intra-class correlation indicates that two experiments from the sample RNA sample but which differ in all other aspects of the study design are highly correlated (median 87.5%). Conversely, two experiments from the same lane, but different RNA samples, etc, show negligible correlation as is expected when technical variation is low.
