Using deep learning to model the hierarchical structure and function of a cell

Embedding a deep-learning model in the known structure of cellular systems yields DCell, a ‘visible’ neural network that can be used to mechanistically interpret genotype–phenotype relationships.