Resumen de Análisis de datos y modelización matemática para la mejora del pronóstico en hematooncología pediátrica
Hematological malignancies are the most frequent cancers in patients between 0 and 19 years, with Acute Lymphoblastic Leukemia and Non-Hodgkin Lymphoma being the most common diagnoses. Chemotherapy schemes initiated in the 1960s have been progressively refined to reach an overall survival nearing 90%, but relapse still constitutes an important cause of disease-related death for this age group. New approaches are needed to identify this portion of patients and to optimize the new therapies that are being developed for tackling relapse.
In this thesis we employ mathematical tools to address these issues. We divide our approach in two separate goals: First, to refine risk assessment by means of machine learning techniques. To this end we leverage flow cytometry data at diagnosis from a large database of pediatric patients. This diagnostic technique holds abundant single-cell information which is not currently employed in risk stratification. Second, to simulate CAR T-cell therapy by means of mathematical models. This therapy has become the most promising alternative for relapse patients but there are still a number of unknowns that preclude its optimal application.
With respect to diagnosis, we show that different characterizations of flow cytometry data are required to uncover its potential prognostic value, pending the availability of high-dimensional omics data at diagnosis and more comprehensive panels of surface proteins. With respect to treatment, we demonstrate that mathematical models adequately describe the dynamics of CAR T-cell therapy and illustrate the crucial role of individual patient data for evaluating hypotheses, testing alternative protocols and identifying biomarkers of response.
