Model-supported Patient Stratification Using Set-based Estimation Methods

Abstract

Stratification of patients into different risk subcategories for disease development plays an important role in medical treatments. It sets the basis for physicians to decide upon personalized interventions. This patient-specific therapy design increasingly becomes supported by mathematical models that describe the underlying disease processes on a detailed molecular level. However, the mathematical description of disease development is challenging. Often the underlying processes act on different time scales. Furthermore, the biomedical data and measurements have different quantities, qualities and uncertainties. New methods are required to address this heterogeneity in the data landscape and to integrate measurements on different time scales in order to extract meaningful information over the disease process. We devise an approach for integrating biological signals for short and long-term molecular processes into a coherent framework. To this end, we combine set-based estimation methods for short-term molecular pathways with classification approaches of long-term disease development. The developed framework is demonstrated by means of IL-6-induced Jak-STAT3 and MAPK trans-signaling.