Cancer is characterized by extreme heterogeneity, with every individual tumor having a distinct pattern of genomic alterations. These patterns affect outcomes and treatment options, but except some well studied alterations we mostly dont understand their specific roles. We analyzed missense somatic mutations in a pan-cancer cohort of 5,989 tumors from 23 projects of The Cancer Genome Atlas (TCGA), mapping them on experimental or predicted models of human proteins and protein-protein complexes. We then searched for enrichment of cancer mutations on specific structural features, such as domains or PPI interfaces and their correlation with disease or treatment outcomes. We identified hundreds of novel domain or interface cancer drivers and found many examples how mutations in different regions in the same gene can have different effects, including patient outcomes. Results of this analysis are available on http://cancer3D.org.