The Moonlight framework for driver gene prediction. Moonlight uses a set of DEGs as input. First, a functional enrichment analysis is carried out to find which of Moonlight’s 101 BPs are overrepresented among the DEGs. Then, a gene regulatory network analysis models how the DEGs are connected with each other through mutual information. Following this step, Moonlight diverges into an expert-based and a machine learning approach. In the next step, an upstream regulatory analysis, the expert-based approach examines the effect of DEGs on user-selected BPs whereas the machine learning approach examines this on all of Moonlight’s BPs. Subsequently, putative tumor suppressors and oncogenes collectively called oncogenic mediators are predicted through a pattern recognition analysis using either patterns (the expert-based approach) or a random forest classifier (the machine learning approach). Finally, a driver mutation analysis analyzes mutations in the cancer patient cohort and categorizes these into drivers and passengers. Those oncogenic mediators containing at least one driver mutation are retained as driver genes.