Fig. 1

Collection of a comprehensive repertoire of cancer and healthy drivers. a Literature review and driver annotation workflow. Expert literature curation of 331 publications led to a repertoire of cancer and healthy drivers in a variety of cancer and non-cancer tissues. Combining multiple data sources, a set of properties and annotations was computed for all these drivers. b Intersection of canonical drivers from three sources [17,18,19] that passed our manual curation. c Classification of canonical cancer drivers in tumor suppressors and oncogenes. Eighty-one cancer drivers had a dual role or could not be classified. d Intersection of canonical and candidate driver genes from 310 sequencing screens. Genes whose driver role had only statistical support were considered candidate cancer drivers. e Intersection between cancer drivers with coding and non-coding alterations. f Level of support for the driver role of 531 cancer genes with non-coding driver alterations only. Level 1 means that the gene was predicted as a driver only in one cancer sequencing screen; levels 2, 3, and 4 mean that it was predicted by two, three, or four screens or that it had experimental support. Experimental support was gathered from the 19 publications reporting non-coding cancer drivers (Additional file 1, Table S1) and from the CNCDatabase [20] and included in vitro and in vivo experiments, modification of gene expression, and survival association. g Proportion of healthy drivers that are also canonical or candidate cancer drivers, classified as canonical and candidate healthy drivers, respectively