Multimodal brain network models for charting the development of brain diseases

Over the past decades, neuroscience has shifted from a localizationist view of brain function toward a network-based framework, recognizing that neurological, psychiatric, and neurodevelopmental disorders often arise from disruptions in the topology and dynamics of distributed brain networks. Traditionally, research has focused on describing how disease alters connectivity, yet little is known about how brain networks themselves constrain the propagation and evolution of pathology. This doctoral work addresses this critical gap, demonstrating that the human connectome shapes disease dynamics across three domains: epilepsy, neuro-oncology, and autism spectrum disorder. Importantly, spatial and temporal aspects of disease progression can be quantified and predicted by integrating computational modeling with network neuroscience approaches. In epilepsy, time-resolved connectivity analyses derived from intracranial EEG, combined with network metrics, revealed how individual network architecture governs seizure initiation, propagation, and failure, identifying critical nodes and dynamical conditions that may inform personalized interventions. In brain tumor patients, pre- and post-operative MRI data were analyzed to characterize how tumors integrate into large-scale brain networks. Deviations in network metrics were shown to potentially serve as biomarkers of tumor infiltration, progression, and functional recovery. Finally, biologically plausible neurocomputational models were used to simulate developmental trajectories of large-scale networks in children with autism, showing how atypical connectivity patterns lead to altered multisensory integration and how targeted interventions can normalize network configurations through Hebbian plasticity mechanisms. Taken together, these findings advance our mechanistic understanding of brain disease as a network phenomenon, moving beyond descriptive approaches to establish a methodological framework, grounded in computational neuroscience, capable of predicting the spatial and temporal patterns of pathology. From a translational perspective, this work lays the foundation for a connectome-informed, patient-specific approach to precision medicine, enabling the characterization of disease mechanisms and informing tailored clinical decision-making, prognosis, and personalized therapeutic interventions.