Computational Exploration of Neural Networks in the Dorsal Cochlear Nucleus

Huebschen, Alan; Stockwell, Daniel

Permalink

https://hdl.handle.net/2142/103593 Copy

Description

Title

Computational Exploration of Neural Networks in the Dorsal Cochlear Nucleus

Author(s)

• Huebschen, Alan
• Stockwell, Daniel

Issue Date

2019

Keyword(s)

Biology

Abstract

Computers have significant application in the realm of neuroscience. The application of computer modeling of biological neural networks allows for inexpensive testing and experimentation. Understandably, there is a demand for accurate models that can serve as preliminary hypothesis testing without the need of live specimens. It is a cheap and ecological technique to further our understanding of how neuronal networks process information. The cochlear nucleus, located in the brain stem, is the first site that processes sound after it leaves the cochlea. The dorsal cochlear nucleus (DCN) has been implicated in deficits seen in both tinnitus and age related hearing loss. Experimentally investigating different changes to these networks using a computer model may lead to insight into these particular disorders. A biophysical model would ideally allow for relatively easy diagnosis of mechanism by altering digital variables to recreate activity observed in pathological individuals. Using a broad software infrastructure, we attempted to create a model of a subset of the DCN. At the lowest level, NEURON (Hines and Carnevale 2001) provides the tools to simulate biophysically accurate models of neurons. It uses a proprietary programming language (hoc) and interfaces with Python. On top of NEURON, Manis and Campagnloa (2018) released the CNModel, a cochlear nucleus simulation framework in Python 3. CNModel is a comprehensive scaffold that makes generating, connecting, and altering cells easy and fast; it utilizes a wealth of physiological data from empirical studies to instantiate objects. We sought to model response properties of neurons when the network was presented with a stimulus. This response is represented as a post stimulus time histogram (PSTH) in which individual responses were grouped based on the time the neuron responded after the stimuli was presented over numerous repetitions of the same test. Some variability is built in to the model but this is still an area that needs improvement as the in silico PSTH has far more defined spiking patterns than the data collected in vivo. At the time of this writing, our current model has been able to come close to replicating the response patterns of DCN pyramidal cells in a young rat; however, the make-up of the network is not physiologically realistic. We have one tenth of the tuberculoventral cells that have been noted in previous research to synapse onto DCN pyramidal cells and completely omit cartwheel cells while driving a stimulus-overlapping inhibitory input into the pyramidal cell with a simulated current clamp. This network makeup is under constant revision with each testing sequence, and we hope to soon implement a more anatomically and physiologically accurate model. The experimentation with stimulus-overlapping inhibitory current clamps has been fundamental in providing clues of how other inhibitory cells might be behaving in live animal models.

Type of Resource

other

Language

en

Permalink

http://hdl.handle.net/2142/103593

Owning Collections