Stimulation-mediated reverse engineering of silent neural networks

Area of Science:

• Neuroscience
• Computational Neuroscience
• Machine Learning

Background:

• Understanding neuronal network connectivity is crucial for brain function.
• Deciphering connections in networks with silent neurons remains a significant challenge.

Purpose of the Study:

• To develop and validate a protocol for reconstructing neuronal network connectivity from simulated silent networks.
• To infer connection weights and predict spike trains with high accuracy using stimulation and supervised learning.

Main Methods:

• Simulated silent neuronal networks with heterogeneous connections and lognormal firing distributions.
• Application of a supervised learning algorithm combined with targeted electrical stimulation.
• Analysis of inferred connection weights and predicted spike train accuracy at single-cell and single-spike levels.

Main Results:

• High-fidelity inference of connection weights.
• Accurate prediction of spike trains at single-spike and single-cell levels.
• Demonstrated improved performance in rat cortical recordings using the developed protocol.

Conclusions:

• The developed protocol effectively reconstructs connectivity in simulated silent neuronal networks.
• Stimulation combined with supervised learning offers a powerful approach for deciphering neuronal connectivity.
• This method has potential applications for both biological and artificial neural networks.