Inherent dynamical vulnerability patterns in human brain connectivity: activity-dependent network diffusion model of neurodegenerative diseases

The combination of advanced brain magnetic resonance imaging (MRI) and complex network ap- proaches has led to very active research on large-scale structural and functional network connectivi- ty in healthy and diseased human brains. Topological analysis of brain networks has established the network-degeneration view for various neurodegenerative diseases (most commonly Alzheimer’s diseases (AD)), showing that the brain structural and functional networks are disrupted and atrophy patterns of various dementias are found to fall into dissociated and dispersed sub-networks. Net- work-degeneration is consistent with the neuropathological evidences of activity-dependent deposi- tion and transsynaptic spread of diseases agents in brain networks. More interestingly, it has been shown very recently that some network patterns in healthy brains could predict partially the atrophy patterns in patients. These observations suggest that there are likely common network-based mecha- nisms for various neurodegenerative diseases, rooted in the inherent vulnerability patterns of normal brain networks, and activity-dependent deposition and diffusion of diseases factors play crucial roles to selectively target the vulnerable sub-networks. Better understanding such inherent vulnerability patterns, especially, identifying possible focal sources of diseases and the dynamical progression from them are of crucial importance for developing sensitive network-based biomarkers for accurate early and differential diagnosis. However, topological analysis is still mainly descriptive and the ac- tivity-dependent deposition and diffusion of diseases agents have not been incorporated simultane- ously into a more integrated dynamical network framework in data analysis or modelling. Therefore the predictive power of several network-based measures is still low. A more elaborated dynamical network approach with a synergy of the heterogeneity in activity-dependent deposition and diffusion of disease agents, both controlled by the underlying brain networks, is desirable in order to signifi- cantly improve the matching between theoretical prediction and experimental observation.

This project proposes to contribute our expertise on dynamical complex networks to this important problem through close collaboration with neuroimaging expert Co-Investigator (Co-I). Different from topological analysis solely, we emphasize on the importance of a trade-off between the com- peting constraints of spatial/metabolic cost and functional requirements in brain networks, which would form hubs and other key nodes with high activity and metabolism to integrate distant func- tional modules and could inevitably build inherent vulnerability in the system. We aim to formulate an activity-dependent reaction-diffusion model calibrated by multimodal brain imaging data to better describe the dynamical evolution of the vulnerability patterns. The model would extend a recent lin- ear and passive network diffusion model with several simple assumptions to capture the essential aspects of activity-dependent deposition and diffusion of disease agents as dynamics-structure co- evolving processes. Compartments of the model will be used to investigate various relationships be- tween the different modals of brain imaging data in young healthy group. Calibrated and driven by multimodal imaging data, the model could self-organize one or several vulnerable focal sources like- ly consisting of different combinations of hubs and other non-hub key nodes with high metabolism in the brain networks. These focal sources then could initiate the progression of vulnerability into dissociated disease patterns as attractors of the nonlinear dynamical network. The dynamical pat- terns of the model will be analysed and compared to disease patterns in MRI data of AD patients provided by the Co-I. We envision that synergizing the heterogeneity in activity and diffusion in the integrative, data-driven model would clearly improve the agreement between theoretical patterns in the model and the measured patterns in patients. With a deeper understanding of the network-based mechanism, this collaborative project will set a foundation for future applications to develop novel biomarkers using the dynamical vulnerable patterns for early and differential diagnosis of AD and other dementias.