An Approximation to the Adaptive Exponential Integrate-and-Fire Neuron Model Allows Fast and Predictive Fitting to...

Area of Science:

• Computational neuroscience
• Systems neuroscience
• Neuronal modeling

Background:

• Large-scale network simulations require computationally efficient yet physiologically valid neuron models.
• Existing models often struggle to predict neuronal responses across diverse input regimes.

Purpose of the Study:

• To develop a high-throughput parameter estimation method for spiking neuron models.
• To ensure models are physiologically valid and predictive across different input conditions.

Main Methods:

• Parameter estimation using closed-form firing rate expressions from an adaptive exponential integrate-and-fire (AdEx) model approximation.
• Fitting models to standard in vitro f-I curves (firing rate vs. input current).
• Testing model predictive accuracy on independent datasets with fluctuating, in vivo-like input currents.

Main Results:

• The fitting procedure is approximately 100 times faster than numerical integration methods.
• Models accurately predicted spike times on test traces across various cell types, cortical layers, and input regimes.
• The method demonstrated good generalization performance, relying solely on firing rate information.

Conclusions:

• A simple, fast, and accurate method for fitting neuron models based on readily available physiological data has been developed.
• This approach enables the creation of computationally tractable and predictive neuron models for large-scale network simulations.
• The method shows promise for efficiently characterizing neuronal dynamics across different brain regions and conditions.