The Shape of Progression: Patterns and Principles in Computational Oncology

This thesis presents a dual approach to computational oncology, bridging data-driven inference of clinical observations with principle-based simulation of biological mechanisms. The work is structured as a methodological portfolio, exploring how different computational frameworks can be utilized to interpret the evolutionary complexity of cancer across varying scales and data types. The first part of this thesis addresses the challenge of interpreting longitudinal biomarkers in clinical practice. Clinical data is frequently sparse, noisy, and irregularly sampled, which often forces a "snapshot" paradigm where patients are stratified based on single measurements at fixed time points. This approach fails to capture the continuous, dynamic nature of disease progression. To address this, we utilize CONNECTOR, a functional data analysis framework specifically designed to recover the underlying biological trajectories from fragmented clinical observations. By treating sequential measurements as continuous functional objects rather than independent data points, the framework identifies shared patterns of evolution within a population. While CONNECTOR is a versatile tool applicable to various longitudinal contexts, this thesis focuses on its application to Minimal Residual Disease (MRD) monitoring in Mantle Cell Lymphoma (MCL). Using data from the FIL-MCL0208 phase III trial, we demonstrate how this workflow uncovers kinetic signatures that correlate with favorable and unfavorable outcomes. To expand the scope of this inference, we also introduce MultiConnector, an extension enabling the joint analysis of multiple longitudinal streams, such as the synchronized evolution of disease in peripheral blood and bone marrow. The second part of the thesis shifts toward a principle-based perspective to investigate the emergence of Intratumor Heterogeneity (ITH). ITH arises from the combined effects of genetic alterations, clonal interactions, and environmental constraints, playing a central role in therapeutic resistance and disease progression. Because these dynamics are often invisible to current diagnostic tools, we present INSITE, a stochastic modeling framework designed to explore the biological rules governing tumor evolution. INSITE integrates genotypic inheritance with phenotype-driven functional traits and resourcemediated competition. Mutational events are associated with functional capabilities—such as altered proliferation, increased mutation rates, access to additional space/resources or enhanced control over shared resources—allowing multiple genotypes to converge on similar phenotypes. The model explicitly tracks subclonal lineages while incorporating environmental constraints that modulate growth. To facilitate reproducibility and exploration, the framework includes an efficient simulation algorithm and an open-source graphical user interface. Using this model, we illustrate how ecological feedbacks shape clonal dynamics over time, supporting an interpretation in which early growth is dominated by stochastic expansion, while later evolution reflects selection for traits that alleviate environmental constraints.