We are excited to share a new preprint, “Learning the mutational landscape of the cancer genome” (Sherman*, Yaari*, Priebe* et al.). This work, a collaboration with Bonnie Berger’s group at MIT, developed a deep-learning model to predict cancer-specific neutral mutation rates at kilobase-scale resolution from epigenomic annotations. Applying this model to the Pan-Cancer Analysis of Whole Genomes (PCAWG) resource identified potential new driver mutations in understudied regions of the genome, including cryptic alternative splice sites, 5′ UTRs, and infrequently-mutated driver genes.