Application of variancePartition to post mortem RNA-seq data of multiple tissues tissues from the GTEx Consortium [2] decouples the influence of multiple biological and technical drivers of expression variation. We analyzed 489 experiments from 103 individuals in 4 tissues (blood, blood vessel, skin and adipose tissue) in order to restrict the analysis to tissues with RNA-seq data for most individuals (Additional file 1: Table S1). Variation across tissues is the major source of variation (median 37.4%) while the technical variables expression batch (2.9%), ischemic time (1.2%), RNA isolation batch (0.4%), and RIN (0.2%) have a moderate effect on expression variation genome-wide (Fig. 5 a). Variation across expression batches is correlated with GC content but to a lesser degree that other datasets (Additional file 1: Figure S6). Cumulatively, these technical variables explain only 4.7% of the total expression variation. Concerns about reliability of RNA-seq data from post mortem samples has been raised due to the potential effects of RNA degradation following cell death [44, 45]. variancePartition analysis indicates that variation in ischemic time has as relatively small effect genome-wide and the fraction of variance it explains is comparable to technical effects, yet the effect varies substantially across genes.
