UCLA

Cancer remains a formidable public health challenge, and identifying effective therapeutic strategies to prevent tumor cell proliferation is paramount to improving patient outcomes. Tumor cells exhibit remarkable phenotypic plasticity, enabling them to assume a diverse range of molecular and phenotypic states, and rapidly develop resistance to therapeutic or environmental stressors. This plasticity, however, presents unique opportunities to identify molecular programs that can be targeted for therapeutic purposes. Therefore, gaining a comprehensive understanding of how clinically relevant anti-cancer agents modulate cell cycle progression is pivotal to uncovering such strategies. In this thesis, we present a suite of computational models that shed light on how drugs modulate the cell cycle, how quantifying drug effects on the cell cycle can inform drug combination recommendations, and how to analyze the heterogeneous response of single cells to cancer therapy. Specifically, Chapter 1 introduces a mathematical model that captures drug-induced dynamical responses, quantified cell cycle phase arrest, and cell death induction rates in cancer cells upon treatment using live-cell microscopy experiments. Leveraging this model, we predict drug combination effects and identify combination treatment strategies that can optimize therapeutic response in cancer, while accounting for specified cell cycle effects. In Chapter 2, we expand the application of this modeling strategy by exploiting a newly introduced simplified experimental assay with fixed cell imaging, thereby broadening the scope of experimental data used for predicting drug combinations with our approach. This chapter also highlights the utility of a mathematical tool to discern general biological patterns within large-scale multi-dimensional data. Finally, in the last chapter, we provide a computational approach to account for phenotypic heterogeneity in drug response observed at the single cell level. We develop a tree-based hidden Markov model that quantifies various drug-induced phenotypic cell states and transition rates between these states resulting from drug-induced cell cycle effects. This approach has potential for uncovering the relationship between molecular states and cellular phenotypes using end-point spatial transcriptomic profiles of cells under treatment. In summary, this work presents a compelling case for how computational models can aid in understanding the effects of anti-cancer agents on the cell cycle and identifying optimal drug combinations. The models presented in this thesis provide an important foundation for further investigations into developing effective therapeutic strategies for cancer treatment.