Abstract One of the greatest challenges in studying the genomic basis of human cancer is distinguishing the few mutations that directly contribute to tumorigenesis (“drivers”) from the many biologically neutral mutations (“passengers”). Existing methods can identify highly recurrent individual mutations or mutated genes, but capturing functional yet infrequent “long-tail” mutations remains challenging. While individually uncommon, these long-tail mutations are present in as many as one-third of cancer patients, and for many of them targeted therapies currently exist. Therefore, improved approaches to identify individual driver mutations in the long tail will facilitate our understanding of their potentially diverse biological and clinical significance. Some of these mutations can be identified as hotspots based on their recurrence on or around the same residue. Others may occur in different regions but are actually close in protein three-dimensional structures. We have developed a novel method that identifies mutations that significantly cluster together within three-dimensional proximity in protein structures, or 3D hotspots. We applied this method to a combined collection of mutation data of 21,000 sequenced tumor samples in 74 cancers. Our analysis confirmed many well-studied 3D hotspots of functional mutations in cancer genes such as KRAS, BRAF, IDH2, SMAD4, FBXW7, SPOP, RHOA, and PTPN11. Most of these genes have well known individual hotspots, but our analysis suggests that additional mutations in other residues of the protein are adjacent in the folded protein structure and may also be oncogenic. We also identified novel 3D hotspots in known cancer genes such as EP300, MAP2K1, and KDR, as well as several other genes of unknown significance that harbor 3D hotspots (e.g., DCC). Most of these hotspots are present in multiple cancers. To explore the functional consequences and potential translational significance of long-tail low-incidence mutations identified by our method, we assessed several MAP2K1 mutations in vitro. MAP2K1 (MEK1) is a critical effector of MAPK signaling, harbors conventional single-codon hotspots in several cancer types, and a number of selective MAPK pathway inhibitors are either FDA-approved or are being investigated in early-phase clinical trials. Our experiments confirmed that these mutations activate MAP2K1 and confer sensitivity to MAP2K1 inhibition. We have provided an implementation of the algorithm via an interactive web resource connecting to cBioPortal for Cancer Genomics (http://cbioportal.org/) for easy interpretation of cancer genomics data in the context of three-dimensional structures. By adding to cBioPortal's already powerful capacity of clinical decision support, our method and analysis provide a useful approach of interpretation, prioritization and extension of biologically significant and clinically actionable mutations in cancer. Citation Format: Jianjiong Gao, Matthew T. Chang, Brooke E. Sylvester, Hannah C. Johnsen, Sizhi P. Gao, S. Onur Sumer, David B. Solit, Barry S. Taylor, Nikolaus Schultz, Chris Sander. Identification of oncogenic mutation hotspots via three-dimensional proximity. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 3606.