Hybrid modeling framework for tumor growth with chemotherapy

• Ano de defesa: 2022
• Autor(a) principal: Naozuka, Gustavo Taiji
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Laboratório Nacional de Computação Científica; Coordenação de Pós-Graduação e Aperfeiçoamento (COPGA); Brasil; LNCC; Programa de Pós-Graduação em Modelagem Computacional
• Palavras-chave em Português: Câncer - Modelos matemáticos; Teoria de controle; Sistemas não-lineares; Sistemas dinâmicos diferenciais; Análise de sensibilidade; CNPQ::CIENCIAS DA SAUDE::MEDICINA::CLINICA MEDICA::CANCEROLOGIA
• Link de acesso: https://tede.lncc.br/handle/tede/362

Cancer is one of the main causes of death in the world, whose mechanisms of origin and growth are not perfectly understood. The multiple spatial and temporal scales of this group of diseases make their understanding even more difficult. In this sense, mathematical and computational modeling is a useful tool that contributes to the understanding of the tumor growth dynamics, as well as to the investigation of the tumor response to different treatment protocols. This Ph.D. dissertation aims to develop a hybrid mathematical and computational model able to characterize the multiple scales present in the growth of solid tumors. We consider that the mechanisms that contribute to the development of cancer act on three scales: tissue, cellular, and molecular. The tissue and molecular scales are described using continuous models, and the cellular scale is represented through a discrete (individual-based) model. In order to investigate different mechanisms of response to chemotherapy, we incorporate the chemotherapeutic drug dispersion in the tumor microenvironment and the signaling dynamics associated with cell cycle control into the hybrid multiscale model. As a significant novelty, this Ph.D. dissertation aims to investigate procedures to obtain the most suitable therapeutic protocol to eliminate the tumor and reduce toxicity in a multiscale context. Due to the hybrid multiscale model complexity, we firstly determine a surrogate model by applying a data-driven approach to the scenario without chemotherapeutic drug dispersion. Specifically, we identify a nonlinear dynamical system by using a combined approach that integrates the Sparse Identification of Nonlinear Dynamics (SINDy) method with a global sensitivity analysis (SA) technique. The proposed SINDy-SA approach is able to discover the most parsimonious tumor growth model that best fits the dynamics informed by the in silico data. By incorporating a control input term associated with the drug concentration in the tumor microenvironment, the identified surrogate model can be investigated from the optimal control point of view. As a preliminary result, we have studied a continuous model composed of drug-sensitive and drug-resistant tumor cells and obtained optimal controls with administration by maximum tolerated dose. The proposed hybrid modeling framework will contribute to the development of effective therapeutic protocols for cancer treatment.