Inferring Neuronal Couplings From Spiking Data Using a Systematic Procedure With a Statistical Criterion.

Area of Science:

• Computational Neuroscience
• Statistical Physics
• Neuroscience

Background:

• Advances in experimental techniques enable analysis of large-scale neuronal networks.
• Inferring neuronal couplings from point process data is crucial for understanding neural dynamics.
• Existing methods may struggle with noise and large datasets.

Purpose of the Study:

• To propose a systematic, objective procedure for pre- and postprocessing point process data.
• To handle neuronal data within a binary statistical model framework (Ising/McCulloch-Pitts).
• To accurately infer neuronal couplings and their signs from complex network activity.

Main Methods:

• Transforming point process data into discrete-time binary data by determining optimal time bin size.
• Utilizing a null hypothesis (independent neuronal firing) and strict criteria for time bin selection.
• Screening relevant couplings by comparing estimates from original data with those from time-randomized datasets to suppress false positives.

Main Results:

• The procedure successfully identifies the presence or absence of synaptic couplings, including their signs.
• Applied to synthetic and in vitro neuronal network spiking data, demonstrating robust performance.
• Effectively suppresses false positive couplings induced by statistical noise.

Conclusions:

• The proposed method offers a reliable approach for inferring neuronal couplings from large-scale point process data.
• It facilitates understanding the physical connections within underlying neural systems.
• The procedure is effective even when employing a simple statistical model for analysis.