Regions in the USA vary widely in many ways, and have large variation in health outcomes. In this work, we investigated how drug prescriptions vary across the USA. The results showed that the country could be partitioned into a patchwork of regional prescription practices, with similar drug use in areas defined by racial composition, health care factors, and wealth. We also found that some areas of the country seem to consistently prefer expensive drugs.

Some Mendelian diseases are associated with increased cancer risk, including hemophilia and other diverse developmental, neurological, and metabolic disorders. We hypothesized that the germline mutations that cause Mendelian disease, and predispose those patients to cancer, have similarity to the somatic alterations that drive cancer development in most patients.

A major goal of my research is to discover effects of drugs using large medical data. In this work, I use neural network methods to create a better way to summarize health care data in order to discover these effects.