Dynamical estimation of neuron and network properties II: Path integral Monte Carlo methods

Area of Science:

• Computational Neuroscience
• Biophysics
• Mathematical Biology

Background:

• Hodgkin-Huxley (HH) models describe neuronal membrane dynamics using nonlinear differential equations for ion channel conductance.
• Estimating parameters and state variables in these models is challenging due to noisy measurements, model errors, and unknown initial conditions.

Purpose of the Study:

• To develop a method for estimating unknown parameters and unobserved state variables of HH models using only voltage observations.
• To characterize biological properties like ion channel complement and density from electrophysiological behavior.

Main Methods:

• Utilized a Monte Carlo numerical approach to directly evaluate the path integral of the joint probability distribution of observed voltage and unobserved model states/parameters.
• Employed noisy intracellular voltage recordings and complex time-varying current stimulation.

Main Results:

• Accurate and precise estimation of HH model parameters and their posterior uncertainty from short (<50 ms) recordings.
• Successful prediction of future neuronal behavior.
• Demonstrated robustness to errors in model specification.

Conclusions:

• The developed method enables direct evaluation of path integrals for HH model parameter estimation.
• This technique accurately characterizes neuronal biophysical properties even with incomplete knowledge of channel expression.
• The approach supports robust model development for diverse biological preparations.