Cancer is fundamentally a disease of the genome. Numerous mutations accumulate in tumors, but only a few specific mutations actually “drive” the growth of cancer cells. We took a data science perspective to distinguish these key mutations by analyzing thousands of human cancer samples across 33 different types of cancer. Through developing novel statistical models that interpret the pattern of mutations observed in cancer, we found new genes and mutations associated with cancer. We found that although particular cancer-associated mutations may occur rarely in patients’ cancers, the overall prevalence of rare cancer-associated mutations suggest they have a critical, underappreciated role in cancer. This may have future implications for precision oncology, where interpretation of a cancer genome will need to be increasingly personalized, since key mutations in a patient’s cancer may have not been previously observed. This work was done in the Rachel Karchin lab (departments of Biomedical Engineering and Oncology and the Institute for Computational Medicine).