Kyung Geun Kim

EECS Department, University of California, Berkeley

Technical Report No. UCB/EECS-2021-229

December 1, 2021

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-229.pdf

Neurons are the most fundamental cells in the brain responsible for processing information. The most essential characteristic that discriminates neurons from other cells comes from their ability to produce electrical signals. The electrical activities are governed by the ion-channels distributed along the membrane of the neurons, making them a critical component to understand the complex neuronal behavior. Compartmental modeling allows us to efficiently model and simulate ion-channels connected to the activities in neurons. Although such models allow meaningful predictions, the quality, as well as the generalizability of the predictions depends heavily on the biophysical accuracy of the models. The accuracy of the model for each neuron recorded from could be improved with an optimization procedure by constraining the model parameters to best fit to the experimental datasets. Depending on both empirical and mathematical aspects of how the optimization procedure is implemented, it is not difficult to arrive at several solutions that fit reasonably well to the experimentally recorded target. However, there exists a single, unique solution that is an accurate description of the neuron recorded from due to the fact that it is a physical system. Current state-of-the-art methods to constraint these model parameters are not yet successful at recovering the unique solution. In this paper, we show that as more conditions are enforced, we could guide the optimization algorithm to eliminate inaccurate solutions and theoretically recover a unique solution that could best represent experimental data. We propose the Conductance based Model Parameter Evaluation (CoMParE) algorithm, which is an algorithmic framework for inferring ion-channel distributions of the neurons recorded from. In order to construct and test the CoMParE algorithm, we first began with Mainen and Sjenowski’s 1996 model of a cortical pyramidal cell with 12 free parameters describing ion channel distribution within the dendritic, somatic, and axonal compartments. With the original parameters of Mainen’s model predefined as the ground truth values, we first showed that the CoMParE algorithm is capable of recovering the ground truth values accurately. Next, we showed that our method could be generalized to other models as well as recover a general point in the parameter space. As a result, the CoMParE algorithm shows a wide potential and capability of fitting biologically detailed models to experimental datasets.

Advisors: Kannan Ramchandran

---

BibTeX citation:

@mastersthesis{Kim:EECS-2021-229,
    Author= {Kim, Kyung Geun},
    Title= {CoMParE: Conductance based Model Parameter Evaluation},
    School= {EECS Department, University of California, Berkeley},
    Year= {2021},
    Month= {Dec},
    Url= {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-229.html},
    Number= {UCB/EECS-2021-229},
    Abstract= {Neurons are the most fundamental cells in the brain responsible for processing information. The most essential characteristic that discriminates neurons from other cells comes from their ability to produce electrical signals. The electrical activities are governed by the ion-channels distributed along the membrane of the neurons, making them a critical component to understand the complex neuronal behavior. Compartmental modeling allows us to efficiently model and simulate ion-channels connected to the activities in neurons. Although such models allow meaningful predictions, the quality, as well as the generalizability of the predictions depends heavily on the biophysical accuracy of the models. The accuracy of the model for each neuron recorded from could be improved with an optimization procedure by constraining the model parameters to best fit to the experimental datasets. Depending on both empirical and mathematical aspects of how the optimization procedure is implemented, it is not difficult to arrive at several solutions that fit reasonably well to the experimentally recorded target. However, there exists a single, unique solution that is an accurate description of the neuron recorded from due to the fact that it is a physical system. Current state-of-the-art methods to constraint these model parameters are not yet successful at recovering the unique solution. In this paper, we show that as more conditions are enforced, we could guide the optimization algorithm to eliminate inaccurate solutions and theoretically recover a unique solution that could best represent experimental data. We propose the Conductance based Model Parameter Evaluation (CoMParE) algorithm, which is an algorithmic framework for inferring ion-channel distributions of the neurons recorded from. In order to construct and test the CoMParE algorithm, we first began with Mainen and Sjenowski’s 1996 model of a cortical pyramidal cell with 12 free parameters describing ion channel distribution within the dendritic, somatic, and axonal compartments. With the original parameters of Mainen’s model predefined as the ground truth values, we first showed that the CoMParE algorithm is capable of recovering the ground truth values accurately. Next, we showed that our method could be generalized to other models as well as recover a general point in the parameter space. As a result, the CoMParE algorithm shows a wide potential and capability of fitting biologically detailed models to experimental datasets.},
}

EndNote citation:

%0 Thesis
%A Kim, Kyung Geun 
%T CoMParE: Conductance based Model Parameter Evaluation
%I EECS Department, University of California, Berkeley
%D 2021
%8 December 1
%@ UCB/EECS-2021-229
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2021/EECS-2021-229.html
%F Kim:EECS-2021-229