Inverse Problems in data-driven multi-scale Systems Medicine: application to cancer physiology

Systems Medicine is an interdisciplinary framework involving reciprocal feedback between clinical investigation and mathematical modeling/analysis. Its aim is to improve the understanding of complex diseases by integrating knowledge and data across multiple levels of biological organization. This Thesis focuses on three inverse problems, arising from three kinds of data and related to cancer physiology, at different scales: tissues, cells, molecules. The general assumption of this piece of research is that cancer is associated toa path ological glucose consumption and, in fact, its functional behavior can be assessed by nuclear medicine experiments using [18F]-fluorodeoxyglucose (FDG) as a radioactive tracer mimicking the glucose properties. At tissue-scale, this Thesis considers the Positron Emission Tomography (PET) imaging technique, and deals with two distinct issues within compartmental analysis. First, this Thesis presents a compartmental approach, referred to as reference tissue model, for the estimation of FDG kinetics inside cancer tissues when the arterial blood input of the system is unknown. Then, this Thesis proposes an efficient and reliable method for recovering the compartmental kinetic parameters for each PET image pixel in the context of parametric imaging, exploiting information on the tissue physiology. Standard models in compartmental analysis assume that phosphorylation and dephosphorylation of FDG occur in the same intracellular cytosolic volume. Advances in cell biochemistry have shown that the appropriate location of dephosphorylation is the endoplasmic reticulum (ER). Therefore, at cell-scale, this Thesis formalizes a biochemically-driven compartmental model accounting for the specific role played by the ER, and applies it to the analysis of in vitro experiments on FDG uptake by cancer cell cultures obtained with a LigandTracer (LT) device. Finally, at molecule-scale, this Thesis provides a preliminary mathematical investigation of a chemical reaction network (CRN), represented by a huge Molecular Interaction Map (MIM), describing the biochemical interactions occurring between signaling proteins in specific pathways within a cancer cell. The main issue addressed in this case is the network parameterization problem, i.e. how to determine the reaction rate coefficients from protein concentration data.