Signals in stochastically generated neurons.

To incorporate variation of neuron shape in neural models, we developed a method of generating a population of realistically shaped neurons. Parameters that characterize a neuron include soma diameters, distances to branch points, fiber diameters, and overall dendritic tree shape and size. Experimentally measured distributions provide a means of treating these morphological parameters as stochastic variables in an algorithm for production of neurons. Stochastically generated neurons shapes were used in a model of hippocampal dentate gyrus granule cells. A large part of the variation of whole neuron input resistance R(N) is due to variation in shape. Membrane resistivity Rm computed from R(N) varies accordingly. Statistics of responses to synaptic activation were computed for different dendritic shapes. Magnitude of response variation depended on synapse location, measurement site, and attribute of response.