NPIN

NPIN classifies neuronal polarity using a node-based machine learning algorithm to assign axonal and dendritic identities from neuronal morphological nodal data for mapping directional signal flow.

Key Features:

• Machine Learning Approach: A node-based machine learning algorithm leverages nodal information to determine neuronal polarity.
• Soma Features: Spatial information from each node to the soma that indicates directionality relative to the cell body.
• Local Features: Morphological details of each node that help distinguish structural components of neurons.
• Spatial Correlations: Integration of spatial correlations between nodal polarities to enhance predictive accuracy, achieving over 96% success even for complex neurons with multiple dendrite/axon clusters.
• Training Dataset and Cross-species Application: Trained on 213 projection neurons from the Drosophila brain and applied to classify neuronal polarity in other insects such as the blowfly, demonstrating adaptability with limited data.

Scientific Applications:

• Neuroscience Research: Mapping signal flows within neural networks to infer how information is processed in insect brains.
• Comparative Neuroanatomy: Comparing neuronal structures and functions across insect species to investigate broader neurobiological principles.

Methodology:

NPIN is trained on nodal data using soma-relative spatial features and local morphological features and integrates spatial correlations between nodal polarities during prediction.