Objective Supervised Machine Learning-Based Classification and Inference of Biological Neuronal Networks.

Area of Science:

• Neuroscience
• Computational Biology
• Machine Learning

Background:

• Classifying biological neuron types and networks is crucial for understanding brain organization and function.
• Existing neuroinformatics approaches face challenges due to limitations in available data, particularly conventional morphological data.

Purpose of the Study:

• To develop a novel objective classification model for biological neuronal morphology and electrical types and their networks.
• To leverage neuronal communication attributes and supervised machine learning for improved classification accuracy.

Main Methods:

• Constructed two open-access computational platforms, Neurpy and Neurgen, using Blue Brain Project models.
• Applied network tomography to cortical neuronal circuits for morphological, topological, and electrical classification.
• Utilized supervised machine learning classifiers (SVM, Decision Trees, Random Forest, ANNs) on simulated data from 10,000 network topology combinations.

Main Results:

• Achieved classification accuracies of up to 70% for neuron types and networks.
• Inferred biological network structures using network tomography with up to 65% accuracy.
• Identified Support Vector Machine (SVM) as a high-performing classifier among those tested.

Conclusions:

• Objective classification of biological networks is feasible using cascaded machine learning methods and neuron communication data.
• This research provides a roadmap for future brain-machine interfaces enabling in vivo objective neuron classification.
• The developed model enhances the understanding of brain structure and offers a novel sensing mechanism.