A mean-field approach to the dynamics of networks of complex neurons, from nonlinear Integrate-and-Fire to Hodgkin–Huxley models

Mallory Carlu, Omar Chehab, Leonardo Dalla Porta, Damien Depannemaecker, Charlotte Héricé, Maciej Jedynak, Elif Köksal Ersöz, Paolo Muratore, Selma Souihel, Cristiano Capone, Yann Zerlaut, Alain Destexhe, Matteo di Volo

TL;DR

This paper introduces a powerful mean-field framework that bridges the gap between the detailed biophysics of individual neurons and the collective dynamics of large neural networks. By developing a method to derive accurate "transfer functions" for complex neuron models, they can predict network behavior across different levels of biological realism, from simplified Integrate-and-Fire to detailed Hodgkin-Huxley models.

What Problem Does This Paper Address?

Understanding the brain requires us to model neural activity at various scales. While detailed biophysical models capture the intricate dynamics of individual neurons, simulating large networks of these models is computationally prohibitive. Conversely, simpler "rate models" offer macroscopic views but often lack biological realism. A key challenge is to create mean-field models that are both computationally tractable and accurately reflect the behavior of complex, conductance-based neurons, especially in the "asynchronous irregular" (AI) firing regimes characteristic of awake cortical activity. This paper aims to bridge this gap by developing a method to derive accurate mean-field descriptions for networks of complex spiking neurons.

Methods

The core of this work lies in developing a "semi-analytic" approach to estimate the transfer function of individual neurons. The transfer function describes how a neuron's output firing rate changes in response to its excitatory and inhibitory inputs. For simple neuron models, this function can be derived analytically. However, for complex models like Hodgkin-Huxley, it's analytically intractable.

The authors' approach involves:

  1. Numerical Simulation: Simulating the behavior of individual complex neurons (AdEx, Hodgkin-Huxley, Morris-Lecar) under various input conditions.
  2. Sub-threshold Statistics: Extracting key statistics of the sub-threshold membrane voltage (mean, standard deviation, autocorrelation time) from these simulations.
  3. Phenomenological Fitting: Using a parameterized formula (a polynomial fit) to relate these sub-threshold statistics to the neuron's output firing rate. This parameterized function serves as the neuron's transfer function.
  4. Mean-Field Formalism: Employing a master equation formalism to derive ordinary differential equations that describe the evolution of the population firing rates and their covariances. This formalism relies on the estimated transfer functions.

This method allows them to obtain biologically realistic mean-field models without needing to simulate every single neuron.

Datasets

Not discussed in the paper. The study relies on numerical simulations of established neuronal models.

Key Insights and Ideas

  • Bridging Microscopic and Macroscopic Scales: The most significant insight is the successful development of a method to derive accurate mean-field models from complex, biophysically detailed neuron models. This provides a powerful tool for studying large-scale neural dynamics.
  • Universality of Transfer Functions: Despite the significant differences in the underlying biophysics of AdEx, Hodgkin-Huxley, and Morris-Lecar neurons, their transfer functions can be made remarkably similar by adjusting neuron parameters. This suggests that the emergent network dynamics might be more universal than the specific neuron models used.
  • Predictive Power for Network Dynamics: The mean-field models accurately predict both the spontaneous activity (average firing rates and their fluctuations) and the response to external stimuli in networks of these complex neurons.
  • Importance of Sub-threshold Dynamics: The approach highlights how the statistics of sub-threshold voltage fluctuations are crucial for determining a neuron's output firing rate and, consequently, the network's collective behavior.

Key Results

  • The developed mean-field formalism accurately predicts the spontaneous firing rates and covariances of excitatory (RS) and inhibitory (FS) neuron populations across AdEx, Hodgkin-Huxley, and Morris-Lecar models.
  • The models show good quantitative agreement with direct network simulations for both spontaneous activity and responses to external stimuli, particularly for relatively slow input signals.
  • The transfer functions derived for different neuron models, when adjusted, show similar shapes, suggesting that network behavior might be largely independent of the specific neuron model details, provided the transfer functions are matched.
  • The method successfully captures phenomena like depolarization block in the Morris-Lecar model, indicating its potential for studying pathological states.

Limitations and Future Work

  • Depolarization Block at High Frequencies: While the method shows promise in predicting depolarization block, its quantitative accuracy at very high firing rates (approaching 500-600 Hz for HH) is limited. The authors suggest further refinement of the transfer function estimation for these regimes.
  • Asynchronous Irregular (AI) Regime: The current formalism is primarily designed for and validated in the AI regime. Its applicability to highly synchronous or bursting activity patterns would require further development.
  • Dendritic Complexity: The current models do not explicitly include dendritic trees, which can significantly influence neuronal integration and network dynamics. Extending the formalism to incorporate dendritic morphology and voltage-dependent currents is a key future direction.
  • Experimental Validation: While the approach is inspired by experimental data, direct validation with intracellular recordings and dynamic-clamp experiments on more complex neuronal structures is a crucial next step.
  • Synaptic Plasticity and Adaptation: Future work could incorporate more complex neuronal dynamics like spike frequency adaptation and synaptic plasticity to model phenomena like slow waves and memory.

Summary updated

Abstract

We present a mean-field formalism able to predict the collective dynamics of large networks of conductance-based interacting spiking neurons. We apply this formalism to several neuronal models, from the simplest Adaptive Exponential Integrate-and-Fire model to the more complex Hodgkin-Huxley and Morris-Lecar models. We show that the resulting mean-field models are capable of predicting the correct spontaneous activity of both excitatory and inhibitory neurons in asynchronous irregular regimes, typical of cortical dynamics. Moreover, it is possible to quantitatively predict the population response to external stimuli in the form of external spike trains. This mean-field formalism therefore provides a paradigm to bridge the scale between population dynamics and the microscopic complexity of the individual cells physiology.<b>NEW & NOTEWORTHY</b> Population models are a powerful mathematical tool to study the dynamics of neuronal networks and to simulate the brain at macroscopic scales. We present a mean-field model capable of quantitatively predicting the temporal dynamics of a network of complex spiking neuronal models, from Integrate-and-Fire to Hodgkin-Huxley, thus linking population models to neurons electrophysiology. This opens a perspective on generating biologically realistic mean-field models from electrophysiological recordings.
BibTeX
@article{carlu2019mean,
  author = {Mallory Carlu and Omar Chehab and Leonardo Dalla Porta and Damien Depannemaecker and Charlotte Héricé and Maciej Jedynak and Elif Köksal Ersöz and Paolo Muratore and Selma Souihel and Cristiano Capone and Yann Zerlaut and Alain Destexhe and Matteo di Volo},
  title = {{A mean-field approach to the dynamics of networks of complex neurons, from nonlinear Integrate-and-Fire to Hodgkin–Huxley models}},
  journal = {Journal of Neurophysiology},
  year = {2019},
  doi = {10.1152/jn.00399.2019},
}
Figure