Accurately fitting biophysical neuron models to experimental voltage data enabled by meta-learning

Area of Science:

• Computational Neuroscience
• Biophysics

Background:

• Neuronal firing properties are determined by ion channels.
• Accurately determining ionic conductances from experimental data is crucial for understanding brain function and neurological disorders.
• This inverse problem has been considered intractable.

Purpose of the Study:

• To develop a method for accurately fitting biophysical neuron models to experimental somatic voltage recordings.
• To enable the determination of ionic conductances from experimental data.

Main Methods:

• Developed a meta-learning algorithm named CoMParE (Computational Meta-Parameter Estimation).
• Applied CoMParE to fit standard and enhanced biophysically detailed neuron models to experimental data.
• Analyzed parameter estimate precision and accuracy.

Main Results:

• CoMParE achieved state-of-the-art reproduction of experimental data.
• Fitting models with enhanced electrophysiological detail further improved data reproduction.
• Meta-learning convexified the objective function loss surface, improving fitting.

Conclusions:

• The inverse problem of determining ionic conductances is tractable using the CoMParE algorithm.
• Highly detailed biophysical models can accurately reproduce experimental data.
• This work advances basic understanding and clinical translation for neurological disorders.