Previous 
Up 
Next
By a "realistic" neuron model we here mean a mathematically formulated model of the main processes influencing the intracellular potential of the cell. This involves formulae describing the motion of different types of ions, primarily sodium, potassium and calcium, as well as methods for computing the electrotonic spread along the dendritic branches of the cell. The class of models we have been using is formulated as a set of coupled ordinary differential equations.
When simulating hundreds or even thousands of interconnected neurons, not necessarily identical, special software tools are needed to handle the large number of parameters involved. We have designed a special object oriented specification language including multiple inheritance for this purpose. A simulator, SWIM, based on this language has been implemented including tools for running the actual simulations on either an ordinary Unix workstation, a CRAY vector-computer or on the CM (Ekeberg et al. 1993).
SWIM first compiles the specification into an internal format suitable for fast access during the numerical simulation. The simulation itself is done using a method specially designed to handle the numerical stiffness introduced by the tight electrical coupling between neighboring parts of the cell. We have found that an implicit method is necessary to achieve numerical stability and that the tree structure of the dendrites makes it possible to solve the implicit equations by Gauss-elimination within reasonable time.
The SWIM simulator has been the basic tool for many of our biological modeling efforts. We have also made SWIM available for other groups abroad working with realistic modeling of neuronal networks.
Next