Two computational neural models : rodent perirhinal cortex and crab cardiac ganglion

Neural engineering research has been rapidly growing in prominence in the past two decades, with 'reverse engineer the brain' listed as one of the 14 grand challenges outlined by the National Academy of Engineering. The computational aspect of reverse engineering includes a study of how both single neurons and networks of neurons integrate diverse signals from both the environment and from within the animal and make complex decisions. Since there are many limitations on the experiments that can be performed in alive or isolated biological systems, there is a need of standalone computational models which can help perform 'in silico' experiments. This dissertation focuses on such 'in silico' neuronal models to predict underlying mechanisms of governing interactions and robustness. The first model investigated is that of a rodent perirhinal cortex area 36 (PRC), and its role in associative memory formation. A large-scale 520 cell biophysical model of the PRC was developed using biological data from the literature. We then used the model to shed light on the mechanisms that support associative memory in the perirhinal network. These analyses revealed that perirhinal associative plasticity is critically dependent on a specific subset of neurons, termed conjunctive cells. When the model network was trained with spatially distributed but coincident neocortical inputs, these conjunctive cells acquired excitatory responses to the paired neocortical inputs and conveyed them to widely distributed perirhinal sites via longitudinal projections. Ablation of conjunctive cells during recall abolished expression of the associative memory. The second model focuses on a model for crab cardiac system consisting of five Large Cells (LC) developed using firsthand biological data. The model is then used to study the features of its underlying oscillation in its membrane potential during a rhythm and to reverse engineer the experimentally discovered phenomenon related to network synchrony. The model predicted multiple mechanisms of compensation to restore network synchrony based on compensatory intrinsic conductances. Finally, a third model, related to the second one, was of an improved three-compartmental biophysical model of an LC that is morphologically realistic and includes provision for inputs from the SCs. To determine viable LC models, maximal conductances in three compartments of an LC are determined by random sampling from a biologically characterized 9D-parameter space, followed by a three-stage rejection protocol that checks for conformity with features in experimental single cell traces. Random LC models that pass the single cell rejection protocol are then incorporated into a network model followed by a final rejection protocol stage. Using disparate experimental data, the study provides hitherto unknown structure-function insights related to the crustacean cardiac ganglion large cell, including the differential roles of active conductances in the three compartments. The novel morphological architecture for the large cell was validated using biological data and used to make predictions about function. A testable prediction related to function was that active conductances, specifically, the persistent sodium current, is required in the neurite to transmit the spike waveforms from the spike initiation zone to the soma. Another pertains to the co-variation of maximal conductances of the persistent sodium current with that of the leak current.