Learning neural connectivity from firing activity: efficient algorithms with provable guarantees on topology.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Graph Theory

Background:

• Neuronal network connectivity is crucial for brain function and information processing.
• Identifying neural network topology is a key challenge in neuroscience.
• Current methods for reconstructing neural networks are often invasive, time-consuming, or lack scalability for large datasets.

Purpose of the Study:

• To develop a scalable method for reconstructing neuronal network topology from neural firing activity.
• To address the limitations of existing techniques in handling large-scale neural data.
• To provide theoretical guarantees for network reconstruction accuracy.

Main Methods:

• Formulating neural network reconstruction as a graph learning problem.
• Developing a scalable learning mechanism to infer connections from neuron firing activities.
• Analyzing the conditions for accurate synaptic connection estimation in Leaky Integrate and Fire (LIf) neuron networks.

Main Results:

• A novel, scalable graph learning algorithm for neural network reconstruction was developed.
• The algorithm accurately identifies underlying synaptic connections in LIf neuron networks.
• Performance was validated using both synthetic benchmark data and real electrophysiological recordings from rat hippocampus.

Conclusions:

• The proposed graph learning approach offers a scalable and effective solution for identifying neuronal network connectivity from firing activity.
• This method advances the field of computational neuroscience by enabling large-scale connectome reconstruction.
• The findings pave the way for more comprehensive studies of brain function and dysfunction.