Project summary/abstract Most Americans past middle age are taking one or more common drugs, and these drugs may have hidden impacts on their subsequent risk of major health outcomes like cancer or Alzheimer's disease. Discovering such drug effects could improve disease prevention and suggest drug repurposing opportunities. One approach to this end is to follow the outcomes of people taking each common drug, as recorded in health records data. While health data is growing in size and detail, it is noisy and incomplete. There is a critical need for new methods to discover drug effects from existing data sources. We propose novel systematic approaches to assess drug-wide association with cancer or Alzheimer's disease by modeling the health record. One innovative concept is our parallel efforts mining independent genetics data. We hypothesize drug side effects can be predicted by analysis of pathological processes shared between the drug's original use and the side effect disease. Analyzing evidence of shared etiology between health conditions and cancer or dementia, we will test this hypothesis. Our efforts to systematically discover drug effects from independent health data and genetics will strengthen our methods. Combining both types of evidence can also allow critical evaluation of putative drug effects on a late-onset disease. While many data-driven studies leave evaluation of the results to future work, we integrate rigorous evaluations into our methods as a means to improve them and strengthen their findings. If successful, our findings c improve clinical management of cancers and neurodegenerative diseases, illnesses of high