AbstractsMedical & Health Science


Despite medicine's rigorous pace of advancement, clinical research remains limited by scalability and portability issues. As we think about the needs of cancer epidemiology, we see the need for multilevel modeling powered by scalable and portable informatics-driven approaches. Of novelty in this dissertation is the `Multilevel Framework for Translational Informatics' that enabled pursuit of a line of scientific inquiry regarding the pharamacogenomics and pharmacoepidemiology of metformin in breast cancer and type 2 diabetes mellitus (T2DM). Metformin is an oral biguanide and is a widely prescribed anti-diabetic medication that is considered to be a first-line treatment for T2DM. While metformin is generally well tolerated it displays wide variation in efficacy and rare adverse reactions; its pharmacogenomics are not clearly understood. Due to the epidemic growth of T2DM in the US and the accumulating evidence highlighting potential repurposing of metformin for cancer prevention and treatment it is imperative to understand molecular mechanisms and clinical impacts of metformin. Further, in order to appropriately separate effects due to metformin and breast cancer from social stress, a known modifier of breast cancer biology, it is necessary to incorporate these characteristics into the model in a way that does not lead to overfitting. To highlight this framework I presented my work in three parts: 1) metformin and insulin pharmacoepidemiology, which as a baseline operated on clinical data only; 2) the modifying impact of socioecological context on breast cancer prevalence, which integrated population measures into clinical context; and finally, 3) translational biomedical informatics of metformin pharmacogenomics, which integrated molecular variation within clinical context. While this work elucidated aspects of metformin pharmacogenomics, it primarily aimed to demonstrate the utility of this framework for integrating multilevel data into future cancer epidemiology and translational biomedical informatics research. As we now see the field of biomedical informatics approaching data mining and data science we see a tantalizing opportunity for utilizing and advancing techniques such as these to power clinical research.