Inferring structure of cortical neuronal networks from activity data: A statistical physics approach

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Understanding the relationship between neuronal network structure and activity is crucial for deciphering brain function and evolution.
• Existing methods struggle to accurately infer complex neuronal network dynamics and plasticity.

Purpose of the Study:

• To develop a novel probabilistic method for inferring effective neuronal network structure from activity.
• To identify neuronal types (excitatory/inhibitory) and predict spiking activity.
• To provide insights into neural plasticity and learning mechanisms.

Main Methods:

• Integration of Bayesian statistics, statistical physics, and machine learning.
• Development of a probabilistic algorithm for network inference.
• Validation using synthetic data, in silico models, and in vitro neuronal networks.

Main Results:

• The method accurately infers effective neuronal network structure and predicts neuronal activity.
• The algorithm successfully identifies excitatory and inhibitory neurons.
• Demonstrated superior performance compared to existing network inference methods.
• Validated across diverse network types (modular and homogeneous).

Conclusions:

• The developed probabilistic method offers a powerful tool for understanding neuronal network dynamics.
• This approach enhances insights into learning processes and neural plasticity.
• Facilitates the development of advanced neuron-based computing circuits.