Development and implementation of novel methodologies to improve Pharmacometrics and Systems Pharmacology analysis

During the past decades Pharmacometrics and Systems Pharmacology (PSP) modelling has emerged as a promising discipline within drug development context. Model-based approaches in drug development involve the integration of pharmacokinetics (PK), pharmacodynamics (PD), disease progression and other relevant information to describe complex biological systems and the action of drugs by computational models. The use of such models can have a major impact during all phases of drug discovery and development and may ultimately result in significant cost reductions for the pharmaceutical industry. Modeling and simulation (M&S) in PSP integrates diverse scientific domains including pharmacology, mathematics, computer science, biostatistics, systems biology, and recently even artificial intelligence is being applied in this field. The diversity of this discipline sometimes results in the challenge that people of different backgrounds do not share the same knowledge about the different aspects governing M&S arena. The present thesis explores the possibility to improve standard PSP modelling by integrating different methodologies and tools that can aid to build a bridge between the different disciplines in order to develop more mechanistic pharmacological models. This thesis is structured as follows: Chapter 1 proposes a qualitative modeling strategy which consist on a computational framework to perform simulations of Boolean networks in the R environment and analyze the result of the perturbations on these networks. This framework called SPIDDOR (from Systems Pharmacology for effIcient Drug Development On R) combines the advantages of the parameter-free nature of logical models while providing a good approximation of the qualitative behavior of pharmacological systems, making the use of Boolean networks in SP more accessible to scientist involved in drug development, especially at its early stages. Additionally, this tool has been used to qualitatively evaluate the results of Boolean network models describing pathogenic mechanisms in the autoimmune diseases systemic lupus erythematosus and inflammatory bowel disease. The publications corresponding to these works are added in the Appendix of the thesis. Chapter 2 proposes an optimization technique known as Optimal Control and its application to a PKPD model for the testosterone effects of triptorelin, a synthetic gonadotropin-releasing hormone analog used to induce chemical castration in prostate cancer patients, with the goal of improving the release characteristics of the drug. As the proposed approach is not circumscribed to just this particular problem, the reader will find a comprehensive description of how the critical aspects of defining control variables and selecting the cost functions and constraints were handled. Chapter 3 presents a computational framework based on a stochastic model known as multitype branching process used to explore the dynamic evolution of heterogeneous tumor cell populations. This framework, which also consist on an R package, is called ACESO (from A Cancer Evolution Simulation Optimizer) and incorporates pharmacokinetics and drug interaction effects into the stochastic model. The aim of this tool is to identify optimum dosing schedules that minimize the risk of developing resistance to anticancer therapies. Finally, in Chapter 4 a semi-mechanistic model describing the time course of several circulating biomarkers in advanced melanoma patients treated with adjuvant high-dose interferon α2b is presented in order to evaluate the dynamics of the tumor markers as prognostic factors of the overall survival and progression-free survival of the patients. This treatment-biomarker-survival model is also coupled to another semimechanistic model describing the side effects of interferon therapy in the absolute neutrophil counts of the patients in order to simultaneously analyze the benefits and toxic effects of this treatment