The evolution of synapse models--from numbers to networks to spaces

We review the evolution of synapse models over the last sixty-five years in terms of the changing paradigms: initially, the synapse was modelled only as part of a neuronal system, both as a number (weight of an edge in a connectionist network) and as a channel in a conductance-based model. With the availability of more structural and kinetic data, it came to be seen as a full-fledged biochemical network, accompanied by a shift from the previous "top-down" approach to a more "bottom-up" network reconstruction. Most recently, the synapse is seen as a geometric 3-dimensional space with various processes driving the dynamics. A particular focus is placed on models of the dopamine synapse and their connections to schizophrenia. The advances of detailed modelling on the synaptic level have highlighted the challenges of integrating the various functional levels, which are tightly coupled with processes on different scales in time and space. On the other hand, this effort will contribute to bridging the currently perceived gap between computational neuroscience and (computational) systems biology.