Biophysically interpretable inference of single neuron dynamics.

Area of Science:

• Computational Neuroscience
• Biophysics

Background:

• Identifying ionic channel contributions to neuronal dynamics is crucial but challenging due to noisy data and ill-posed inverse problems.
• Existing methods struggle with the complexity and nonlinearities of neural electrical activity.

Purpose of the Study:

• To develop a biophysically interpretable, learning-based strategy for data-driven inference of neuronal dynamics.
• To approximate membrane potential dynamics as a weighted sum of ionic currents using two novel optimization frameworks.

Main Methods:

• Two optimization frameworks were developed to learn and approximate neural dynamics from voltage trajectories.
• Ionic currents were represented parametrically or as linear combinations of basis functions, using channel kinetics libraries.
• A linear optimization problem was solved to infer weights for approximating membrane potential dynamics.

Main Results:

• The first strategy recovers channel conductances and reversal potentials.
• The second strategy infers active channels and gating variable trajectories, enabling biophysically salient inference.
• The method efficiently infers complex nonlinear neural dynamics from noisy membrane potential recordings across various timescales.

Conclusions:

• The proposed data-driven approach offers a powerful tool for understanding neuronal function by dissecting ionic current contributions.
• This method addresses the mathematical challenges of inferring channel dynamics from limited experimental data.
• The findings advance the field of experimental neuroscience by providing a robust strategy for neural dynamics inference.