Deciphering the genetic code of neuronal type connectivity through bilinear modeling.

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Genomics

Background:

• Understanding neuronal connections is key to brain function, but genetic determinants of specific neuronal type connectivity remain elusive.
• Existing methods face challenges in deciphering the genetic basis of precise synaptic connections between different neuron types.

Purpose of the Study:

• To develop and validate a novel computational approach for predicting neuronal connectivity based on gene expression data.
• To identify the genetic underpinnings that govern specific synaptic connections between neuronal types.

Main Methods:

• A novel bilinear modeling approach, inspired by recommendation systems, was employed.
• Gene expression data from single-cell transcriptomics of presynaptic and postsynaptic neurons were transformed into a covariance matrix.
• This covariance matrix was optimized to mirror a known anatomical connectivity matrix derived from connectomic data.

Main Results:

• The bilinear model demonstrated performance comparable to or better than the Spatial Connectome Model (SCM) in reconstructing electrical synaptic connectivity in *Caenorhabditis elegans*.
• The model identified all genetic interactions found by SCM and suggested additional ones.
• In mouse retinal neurons, the model successfully recapitulated known connectivity motifs and identified unique genetic signatures, including genes involved in cell adhesion and synapse formation.

Conclusions:

• The developed bilinear model offers an innovative computational strategy for decoding the genetic programming of neuronal type connectivity.
• This approach sets a new benchmark for analyzing synaptic connections using single-cell transcriptomics.
• The findings pave the way for mechanistic studies of neural circuit assembly and genetic manipulation of circuit wiring.