Using Molecular and Clinical Data to Stratify Cancer Patients for Precision Medicine

Cancers are heterogeneous across different individuals. Insights derived from clinical and/or molecular data could be used to develop robust patient stratification models to tailor treatments for each individual patient, in order to improve patient outcome and reduce deleterious side effects. My thesis research mainly focused on using computational methods to understand the biological/clinical differences between disease subgroups and their clinical implications in two diseases, lung cancer and germ cell tumor. A comprehensive analysis using The Cancer Genomic Atlas (TCGA) lung adenocarcinoma datasets showed that FOXM1 was likely to play an important role in the morphology differences among different morphological subgroups in invasive lung adenocarcinoma. In collaboration with the Malignant Germ Cell International Consortium (MaGIC), I developed the MaGIC data dictionary as a uniform data standard to build a germ cell tumor data commons. The MaGIC data commons was then used to harmonize and integrate the patient and genomic data from both MaGIC and the public domain. Concurrently, I also developed a prognostic model for pediatric extracranial germ cell tumor using the integrated dataset, identifying older age, higher disease stage and extragonadal site as adverse prognostic factors. The model was evaluated in an independent dataset of data combined from Brazilian and French clinical trials.