I remember going to a really interesting talk in which the proposal was to represent the transcriptome as a probabilistic generative model. That felt like getting something close to an appreciation of the complexity.