The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Data assimilation for conductance-based neuronal models
Author:	Moye, Matthew
View Online:	njit-etd2020-015 (xv, 173 pages ~ 35.5 MB pdf)
Department:	Department of Mathematical Sciences
Degree:	Doctor of Philosophy
Program:	Mathematical Sciences
Document Type:	Dissertation
Advisory Committee:	Diekman, Casey (Committee chair) Rotstein, Horacio G. (Committee member) Matveev, Victor Victorovich (Committee member) Moore, Richard O. (Committee member) Sauer, Tim (Committee member)
Date:	2020-05
Keywords:	Computational neuroscience Data assimilation Neuroscience Optimization Parameter estimation
Availability:	Unrestricted
Abstract:	This dissertation illustrates the use of data assimilation algorithms to estimate unobserved variables and unknown parameters of conductance-based neuronal models. Modern data assimilation (DA) techniques are widely used in climate science and weather prediction, but have only recently begun to be applied in neuroscience. The two main classes of DA techniques are sequential methods and variational methods. Throughout this work, twin experiments, where the data is synthetically generated from output of the model, are used to validate use of these techniques for conductance-based models observing only the voltage trace. In Chapter 1, these techniques are described in detail and the estimation problem for conductance-based neuron models is derived. In Chapter 2, these techniques are applied to a minimal conductance-based model, the Morris-Lecar model. This model exhibits qualitatively different types of neuronal excitability due to changes in the underlying bifurcation structure and it is shown that the DA methods can identify parameter sets that produce the correct bifurcation structure even with initial parameter guesses that correspond to a different excitability regime. This demonstrates the ability of DA techniques to perform nonlinear state and parameter estimation, and introduces the geometric structure of inferred models as a novel qualitative measure of estimation success. Chapter 3 extends the ideas of variational data assimilation to include a control term to relax the problem further in a process that is referred to as nudging from the geoscience community. The nudged 4D-Var is applied to twin experiments from a more complex, Hodgkin-Huxley-type two-compartment model for various time-sampling strategies. This controlled 4D-Var with nonuniform time-samplings is then applied to voltage traces from current-clamp recordings of suprachiasmatic nucleus neurons in diurnal rodents to improve upon our understanding of the driving forces in circadian (~24) rhythms of electrical activity. In Chapter 4 the complementary strengths of 4D-Var and UKF are leveraged to create a two-stage algorithm that uses 4D-Var to estimate fast timescale parameters and UKF for slow timescale parameters. This coupled approach is applied to data from a conductance-based model of neuronal bursting with distinctive slow and fast time-scales present in the dynamics. In Chapter 5, the ideas of identifiability and sensitivity are introduced. The Morris-Lecar model and a subset of its parameters are shown to be identifiable through the use of numerical techniques. Chapter 6 frames the selection of stimulus waveforms to inject into neurons during patch-clamp recordings as an optimal experimental design problem. Results on the optimal stimulus waveforms for improving the identifiability of parameters for a Hodgkin-Huxley-type model are presented. Chapter 7 shows the preliminary application of data assimilation for voltage-clamp, rather than current-clamp, data and expands on voltage-clamp principles to formulate a reduced assimilation problem driven by the observed voltage. Concluding thoughts are given in Chapter 8.