Network structure and function in the input stage of the cerebellar cortex

It has long been recognised that neuronal networks are complex systems, whose dynamics depend on the properties of the individual synapses and neurons and the way in which they are interconnected. However, establishing clear links between network structure and function has proven difficult. To address this question I applied tools and techniques from computational neuroscience, neuroinformatics, information theory, machine learning, spatial point process theory and network theory, deploying them on a suitable HPC infrastructure where appropriate. Moreover, access to electrophysiological and anatomical data enabled me to develop biologically accurate models and to compare my theoretical predictions with analyses of raw data. In this work, I focused on the granule cell layer (GCL), the input stage of the cerebellar cortex. The GCL is particularly well suited to this type of analysis, as its structural characteristics are comparatively regular, well known and conserved across animal species, and several of its basic functions are relatively well understood. I showed that the synaptic connectivity in simple feed forward networks like the GCL governs the trade-off between information transmission and sparsification of incoming signals. This suggests a link between the functional requirements for the network and the strong evolutionary conservation of the anatomy of the cerebellar GCL. Furthermore, I investigated how the geometry of the GCL interacts with the spatial constraints of synaptic connectivity and gives rise to the statistical features of the chemically and electrically coupled networks formed by mossy fibres, granule cells and Golgi cells. Finally, I studied the influence of the spatial structure of the Golgi cell network on the robustness of the synchronous activity state it can support.