Zheng, Youwei (2015) Contributions to Models of Single Neuron Computation in Striatum Cortex. PhD thesis, Institute of Computer Science, University of Rostock.
Full text not available from this repository.Abstract
Single neurons are bioelectrical transformers that continuously convert input spike trains encoding sensory perception alongside internal states of the brain into output spike series. The essence of this computational transformation is nonlinearity, realized by a cascade of nonlinear components within the neuronal circuits: dendrite, spine, synapse and synaptic plasticity. A deeper understanding is required of how a single neuron utilizes its nonlinear subcellular computational devices at different neuroanatomical scales to generate complex neuronaldynamics. The results presented in this thesis focus on single neuron computation. In particular, compartmental models of cortex and striatum are accurately formulated and firmly grounded in the experimental reality of electrophysiology to address two questions: i) how striatal projection neurons implement location-dependent dendritic integration to carry out association-basedcomputation. A new multi-compartmental model is introduced to replicate the regenerative characteristics of distal dendrites observed in the experiment. By applying the model to a new set of stimulation protocols,I find that single neuron’s richness in associative information processing comes from the interplay between proximal and distal dendrites. This behavior is governed by two cellular anatomical ingredients which are the delicate tapering of single dendritic branch and the length of spine neck; ii) how cortical pyramidal neurons strategically exploit the type and location of synaptic contacts to enrich its computational capacities. A new model is validated by the glutamate uncaging experiment that reveals the difference between evoked EPSPs by axo-spinous and axo-shaft synapses. The model prediction demonstrates that distal axo-shaft synapses can gate non linear dendritic computing with higher threshold and drive somatic potential with higher gain. The results shade light on the preserved functional role of subcellular components on neuronal computation across different brain regions. In another separate investigative pursuit, the emphasis is on addressing the question: how point neuron models respond to converging presynaptic inputs to form synaptic patterns, given diverse configurations of input statistics and various kinds of learning rules which are known as spike-timing-dependent plasticity (STDP)? I characterize the potential effects of action potential (AP) dynamics on STDP by exploring a new phenomenological model that incorporates an AP-dependent learning window. The simulation indicates that AP duration is another key factor for insensitizing the postsynaptic neural firing and for controlling the shape of synaptic weight distribution. The models developed in this thesis are not designed to accomplish good performance in model parameter estimation or either modeling formalism, but to predict particular aspects of single neuron computation that involves re-assembling nonlinear fundamental units of neuronal information processing. The results provide strong and testable predictions, which if experimentally validated, could offer new insights into the functional roles of elementary computational units of the brain such as shaft synapses that have not yet been investigated.
| Item Type: | Thesis (PhD) |
|---|
