Neural System Identification With Spike-Triggered Non-Negative Matrix Factorization

Area of Science:

• Neuroscience
• Computational Neuroscience
• Vision Science

Background:

• Neuronal circuits exhibit complex connectivity patterns, even in simpler systems like the retina.
• Retinal ganglion cells (GCs) integrate excitatory inputs to generate action potentials (spikes).
• Systematic analytical methods are crucial for understanding neuronal circuit structure.

Purpose of the Study:

• To extend the applicability of the spike-triggered non-negative matrix factorization (STNMF) method.
• To utilize retinal ganglion cells (GCs) as a model system for circuit analysis.
• To demonstrate STNMF's capability in dissecting neuronal circuit components.

Main Methods:

• Application of the spike-triggered non-negative matrix factorization (STNMF) method to retinal ganglion cells (GCs).
• Analysis of GC spike data to infer properties of presynaptic bipolar cells (BCs).

Main Results:

• STNMF successfully identified computational properties of upstream bipolar cells (BCs), including spatial receptive fields and temporal filters.
• The method accurately recovered synaptic connection strengths from the STNMF weight matrix.
• STNMF demonstrated the ability to segregate GC spikes, attributing subsets to individual presynaptic BCs.

Conclusions:

• The extended STNMF method is a powerful tool for deciphering the structure and function of neuronal circuits.
• STNMF provides insights into the contributions of individual presynaptic neurons to postsynaptic cell activity.
• This approach advances the systematic analysis of neural computation in the retina.