On the development and interpretation of parameter manifolds for biophysically robust compartmental models of CA3 hippocampal neurons

Electrical spiking behaviors in neurons are governed by a class of membrane-spanning proteins known as voltage-gated and/or calcium-dependent (VGCD) channels. Neuroscientists would like to understand how neurons encode information in these spiking behaviors and how VGCD channels contribute to this information processing. We don't have the ability to perform exact measurements or manipulations of VGCD channel distributions experimentally, so in this dissertation I developed a methodology and tools to obtain and interpret these ranges computationally. Currently, investigators studying compartmental models set their system parameters by an iterative, trial-and-error process. Proceeding in this manner can be as inefficient and can lead to the inadvertent selection of mathematical singularities. We would like to know ALL the parameter sets that produce appropriate behaviors because such a manifold, or boundary enclosing a volume of solutions, captures the reality of biology-like diversity.;I verified the accuracy of the simulations and optimized simulator performance in order to maximize the size of the search space we could explore within a fixed allocation of computational resources. I custom-designed an evolutionary algorithm (EA) to explore the 114 dimensional space of a hippocampal CA3 neuron model for solutions that corresponded well with experimentally determined spatiotemporal behaviors of this class of neuron. The resulting channel distribution manifolds corresponded well with experimental data and allowed predictions regarding the contributions of various channel types to the behavior of the model neuron. This research has developed and validated a technique that will aid in the development of biophysically sophisticated neuronal models when insufficient quantitative, experimental information is available to specify system parameters. This technique promises to be more efficient and reliable than current parameter setting methods. The programs developed as part of this research automated the prediction of functional spatial distributions of VGCD channels in a high dimensional space, a problem hindering the development of single neuron models for the past forty years. The contribution of this dissertation to science is uniquely characterized by the automated production and interpretive paradigm of a MANIFOLD approach to computational modeling.