Ising-like model replicating time-averaged spiking behaviour of in vitro neuronal networks

Area of Science:

• Computational Neuroscience
• Statistical Physics
• Complex Systems

Background:

• Neuronal networks exhibit complex activity patterns.
• Understanding emergent network behavior from individual neuron dynamics is challenging.

Purpose of the Study:

• To develop an inverse binary Ising-like model for analyzing in vitro neuronal network activity.
• To identify local fields and interaction couplings governing network behavior.
• To investigate the thermodynamic properties and criticality of neuronal networks.

Main Methods:

• Analysis of time-averaged experimental data from in vitro neuronal networks.
• Application of the Pairwise Maximum-Entropy method.
• Inference of an inverse binary Ising-like model.

Main Results:

• The model successfully reproduced average neuron activities and pairwise correlations.
• Local fields were found to store specific neuron information, while interactions were generally symmetric.
• Third-order neuron correlations were accurately captured by the derived model.
• The model exhibited second-order phase transition characteristics.

Conclusions:

• The derived Ising-like model provides a powerful framework for understanding neuronal network dynamics.
• Neuronal network behavior may exhibit criticality analogous to physical systems.
• The findings support a thermodynamic interpretation of neuronal network function and criticality.