Scalable multi-omics data-driven modeling of metabolism: a systems approach to simulate metabolic reprogramming

Biological systems play a crucial role in addressing challenges across diverse fields, including industrial optimization, environmental sustainability, and medical advancements such as cancer diagnosis and treatment. Their complexity, hierarchical regulation, and adaptive nature make them difficult to analyse, requiring an integrative approach to predict system behaviour and design effective interventions. Systems biology and mathematical modelling have emerged as essential methodologies for translating biological knowledge into quantitative frameworks. These models enable the simulation, analysis, and prediction of system responses under varying conditions. However, the inherent complexity of biological networks, coupled with computational and informational constraints, poses significant challenges for modelling at a detailed level. Recent advancements in high-throughput experimental technologies have produced vast multi-omics datasets, capturing molecular states across multiple layers. Integrating these heterogeneous data provides insights into underlying mechanisms, optimization strategies, therapeutic targets, and system-level properties. Nevertheless, this integration introduces methodological challenges, including data heterogeneity, noise, and scalability requirements. This thesis contributes to biological systems research by integrating multi-omics data and mathematical modelling and an a priori network reduction. The first part focuses on mechanistic kinetic modelling of oxidative stress responses in yeast, employing transcriptomic, metabolomic, and fluxomic data within a linlog kinetic framework. The second part applies machine learning and pathway enrichment approaches to multi-omics datasets for subtyping and early diagnosis of cancer, identifying molecular markers that characterize personalized disease states. By combining mechanistic and data-driven approaches, this work contributes to bridging the gap between molecular-level understanding and predictive modelling, enhancing applications in industrial biotechnology and personalized medicine.