Object-oriented reaction-diffusion modeling in the neuron simulator

Recent progress in neurobiological genomics and proteomics allows us to consider not only how neurons combine information electrically, but also how they couple electrical with chemical signalling along dendrites and axons. A challenge in neurochemical modeling arises from the large lengths of these neuronal processes, which limits the application of biosimulation methods commonly used in bacteria and in more compact eukaryotic cells. We have developed a first version reaction-diffusion (RxD) tool in the NEURON simulation environment that uses several methods to make the RxD calculations of large neurons tractable. Our approach is object-oriented, allowing the option to set up a set of reactions as a RxD object either in the whole tree, or in any subset of dendritic or axonal sections. Each RxD object is integrated separately but can interact with both the electrical integration and with any other chemical integration defined across the same or overlapping dendritic regions. Reaction schemes are defined using the National Biomedical Resource Simulation language utilized in NEURON, permitting translation from standard biochemical markup languages. Segregation into objects permits different species to be integrated on different grids depending on spatial stiffness, steepness of gradient and the rapidity of local alterations based on the associated reaction schemes. Our initial implementation supports simulation of one-dimensional diffusion. This permits integration of substances across long distances that would otherwise be computationally prohibitive. These RxD-1 objects will coexist with higher dimensional integrations that will simultaneously allow detailed simulation in a more restricted region of the cell. Under specialized circumstances, substances can directly pass from one RxD object to another, for example when passing from 1-D in dendrite to 2-D or 3-D in soma. More generally, they will be able to alter sink or source fluxes at particular locations, which may be remote in space and time. This will, for example, permit a reaction or interaction at the nucleus to have a remote, delayed effect that involves fluxes or reaction rates at synapses. This notion of delayed, remote signaling is immediately analogous to the simulation techniques utilized for synaptic interactions between neurons in neuronal networks. Providing object-oriented segmentation of the overall simulation allows appropriate scales and techniques to be utilized for different aspects of the multiscale problem. This will permit exploration of how various chemical signaling systems combine with or complement electrical signaling to transform information in the cell.