Mark Gerstein and colleagues report a new method, FunSeq, to facilitate the analysis and functional prioritization of noncoding variation in cancer genomics studies (Science 342, 1235587, 2013). They analyzed patterns of selection across noncoding regions using data from the 1000 Genomes Project Phase I data set that comprises genome sequences from 1,092 individuals. They used enrichment of rare SNPs as an estimate of purifying selection and further subdivided noncoding regions into 677 specific categories, finding that 102 of these categories showed significant evidence of selective constraint. The highest levels of negative selection were found in regions they term 'sensitive' and 'ultrasensitive', which have, respectively, a 40- and 400-fold enrichment in disease-causing mutations annotated in the Human Gene Mutation Database. Variants in regions with high connectivity (hubs) in protein-protein interaction or regulatory networks showed higher selective constraint. To demonstrate the usefulness of FunSeq in the analysis of cancer genomes to prioritize candidate driver mutations, the authors examined a data set including whole-genome sequences from 64 prostate, 21 breast and 3 medulloblastoma tumors, identifying 98 noncoding candidate drivers. FunSeq, which was also shown to be useful in identifying potentially deleterious variants in personal genome sequences, has been made publicly available as an automated web tool at http://funseq.gersteinlab.org/.