Data-driven Spiking Neural Network Models that Refine Theories of Hippocampal Function

Computational modeling complements and expands the benefits of animal experimentation. Hippocampome.org is a key resource for modeling neural activity exhibited by the hippocampal formation. My work addresses a gap in knowledge, namely how to find and use experimental work to increase the realism of hippocampal spiking neural network simulations. The overarching aim of my doctoral research was to create more Hippocampome.org resources and to use those resources to produce a spatial navigation simulation that closely matches real cell activities. I created a literature review and knowledge base of spiking neural networks in the hippocampal formation named Cognome. Findings included in Cognome, as well as several additional Hippocampome.org properties, informed the use of methods in the creation of a spatial navigation simulation. The simulation’s parameters included the use of results from research projects I worked on such as identifying synaptic probabilities and modeling, the creation of CARLsim 6, and measurements of grid cell firing metrics. I also contributed to research on neuron-type specific oscillatory phase locking in vivo and generated preliminary results of my spatial navigation simulation for Hippocampome.org’s version 2.0 release. Results of my work include several additions to Hippocampome.org resources, such as front- and back-end development of multiple functionalities, including browsable tables, display of experimental evidence, and online computational tools. My work on Cognome facilitates researchers’ abilities to find categorized knowledge on their cognitive functions of interest, as well as accompanying statistics. My work on a spatial navigation simulation adds to knowledge about the subject, for example, creating testable predictions about neural properties that generate characteristic grid cell firing. The open source release of this work makes it a beneficial tool for more computational researchers. The simulation reproduces well an assortment of neural firing activities of real cells. The work thus serves as an example for the research community of how Hippocampome.org resources can be integrated into popular network-level algorithms and theories, and how the resultant biological realism can facilitate comparisons with animal experimentation results.

Author

• Sutton, Nate

• Biophysics
• Computational modeling
• Hippocampal Formation
• Neural Networks
• Neuron Types
• Rodents

Identifier

hdl:1920/14333

OAI identifier oai:identifier

oai:MARS:1920/14333