Computational Inference of Synaptic Polarities in Neuronal Networks

Area of Science:

• Neuroscience
• Computational Biology
• Systems Biology

Background:

• Synaptic polarity (inhibitory/excitatory) is fundamental to neural circuit function but challenging to map.
• Understanding synaptic polarity is key to deciphering complex brain functions and building accurate brain models.

Purpose of the Study:

• To computationally infer synaptic polarity using the Caenorhabditis elegans connectome.
• To develop and test three distinct computational approaches for predicting synaptic polarity based on available data.

Main Methods:

• Integration of neurotransmitter (NT) and receptor (R) gene expression data with the connectome model (CM) and wiring rules.
• Training a spatial connectome model using known polarities and gene expression data.
• Employing network sign prediction for polarity inference without prior expression or wiring rules.

Main Results:

• The connectome model resolved 356 additional synaptic polarities.
• The spatial connectome model inferred polarity for 81% of connections with high precision and identified 147 new polarities.
• A generalized CM demonstrated high performance in polarity prediction without prior data.

Conclusions:

• Computational inference of synaptic polarity is feasible using connectome data and gene expression.
• The developed methods significantly expand the mapping of synaptic polarities in neural circuits.
• These findings advance the creation of more comprehensive and realistic models of brain function.