Machine learning and complex network analysis of drug effects on neuronal microelectrode biosensor data.

Area of Science:

• Neuroscience
• Computational Biology
• Pharmacology

Background:

• Biosensors, like microelectrode arrays, are crucial for in vitro neuronal activity recording.
• Studying neuroactive substances requires advanced analytical methods for complex biosensor data.

Purpose of the Study:

• To develop and validate a machine learning workflow for analyzing drug-induced changes in neuronal biosensor data.
• To apply complex network measures from graph theory to characterize pharmacological effects on neuronal networks.

Main Methods:

• Utilized microelectrode array recordings of neuronal networks exposed to bicuculline (a GABA receptor antagonist).
• Integrated network-based features with synchrony analysis, optimizing preprocessing parameters (e.g., spike train bin sizes, segmentation window sizes, correlation methods).
• Employed machine learning for classification and Shapley Additive Explanations for feature interpretation.

Main Results:

• Achieved high classification accuracy (Area Under the Curve up to 90%) in detecting drug-induced network alterations.
• Identified significant reductions in network complexity and segregation, characteristic of bicuculline-induced epileptiform activity.
• Demonstrated the workflow's ability to detect and characterize pharmacological effects on neuronal networks.

Conclusions:

• The developed machine learning workflow effectively analyzes neuronal biosensor data to detect neuroactive compound effects.
• This framework offers a broadly applicable tool for identifying both strong and subtle network alterations in neuropharmacology.
• The methodology shows potential for advancing biosensor applications in drug discovery and understanding neurological disorders.