Discovering Effective Connectivity in Neural Circuits: Analysis Based on Machine Learning Methodology

Area of Science:

• Neuroscience
• Computational Neuroscience
• Artificial Intelligence

Background:

• Multielectrode array technology generates vast neural data, necessitating efficient analysis tools.
• Classical statistical methods struggle with complex, high-volume spike train data.
• Existing sophisticated tools are computationally expensive and less accessible.

Purpose of the Study:

• To evaluate the efficacy of AI algorithms for analyzing effective neuronal connectivity.
• To assess the C5.0 algorithm's capability in processing simulated and biological spike train data.
• To explore optimized C5.0 processes (combinatory, iterative, recursive) for enhanced performance.

Main Methods:

• Utilized the C5.0 decision tree algorithm for analyzing spike train data.
• Employed simulated neuronal circuit data with known connectivity for initial testing.
• Applied iterative and recursive C5.0 processes to both simulated and biological datasets.
• Tested on a reduced dataset from *in vitro* mouse spinal cord recordings.

Main Results:

• C5.0 successfully identified monosynaptically connected neurons in simulated datasets within a single run.
• Iterative and recursive C5.0 processes identified both monosynaptic and disynaptic connections under optimal conditions.
• The algorithm provided valuable insights into monosynaptic connections within the biological dataset.

Conclusions:

• AI algorithms, particularly C5.0, offer a computationally efficient approach to studying effective neuronal connectivity.
• The developed C5.0-based methods demonstrate significant potential for analyzing complex neural network structures.
• This study provides a strong proof of concept for AI-driven neural connectivity analysis.