Multiscale Modeling Of Cell Fate Switching To Predict Patient-Specific Response To Combination Cancer Therapy
All cells in the human body share the same DNA sequence, but differ in their functional identity, guided by a wide array of regulatory mechanisms controlling cellular lineage commitment and encoded in the unique epigenome of each cell type. Recent experimental studies with induced pluripotent stem cells have allowed researchers to investigate the dynamic nature of cell identity and relationships between gene regulation and differentiation. These studies have major implications for our understanding of not only human development, but also disease as cancers, and some neurological diseases, arise in part due to inappropriate persistence of cells in immature differentiation states. These studies have proliferated massive multi-omics databases as next generation sequencing (NGS) technologies are applied to extensively profile stem cells, in vitro differentiated cell populations, and cancer patient cohorts. As these data accumulate, important unanswered questions remain, including to what extent do physical states of genes change in development and disease, and how do these changes meaningful alter cell signaling pathways and clinically impact individual patients. To address these questions, new computational tools are needed to 1) rigorously assess epigenomic state changes captured with NGS modalities in the course of lineage specification, and to 2) integrate models of patient (epi-)genomic states with those of disease-associated cell signaling pathways and clinical outcomes. This thesis describes the development and applications of computational methods to help address these needs. First described is a statistical tool for classifying long-range looping interactions that change across developmental models and disease states from data captured with NGS technologies. Its application is demonstrated for study of chromatin looping state changes in the course of neural lineage commitment and neuronal activation. Then a multiscale framework is described that integrates patient (epi-)genomic profiles with mechanistic models of signaling pathways critical to decision making to enter tumorigenic states. The framework is demonstrated in a clinical setting to predict patient-specific responses to different specific treatment combinations in nephroblastoma. Such methods have great potential to advance our understanding of the determinants of cellular identity and its loss in cancer, and in turn our ability to personalize patient care.