Analyzing functional connectivity using a network likelihood model of ensemble neural spiking activity.

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Data Analysis

Background:

• Understanding neural communication requires analyzing spike train dependencies.
• Current methods often focus on pairwise neuron interactions.
• Maximum likelihood methods offer a framework for simultaneous multi-neuron analysis.

Purpose of the Study:

• To develop a discrete-time version of a network likelihood model for analyzing ensemble neural activity.
• To improve computational efficiency and provide principled methods for parameter estimation and model convergence.
• To validate the algorithm's performance on simulated and real neural data.

Main Methods:

• Developed a discrete-time estimation algorithm based on a network likelihood model with multiplicative conditional intensity functions.
• Incorporated an efficient computational strategy, a principled method for initial value computation, and a principled stopping criterion.
• Applied the algorithm to analyze simulated neural spike trains and spike trains from rat hippocampal place cells.

Main Results:

• The algorithm successfully recovered correct connectivity matrices and interaction parameters in simulated data.
• Analysis of rat hippocampal place cell data revealed a connectivity matrix consistent with neuronal receptive field overlap.
• Demonstrated accurate identification of network structure and interaction parameters.

Conclusions:

• The discrete-time network likelihood algorithm is an efficient tool for analyzing ensemble spiking activity.
• The method accurately captures pairwise and higher-order dependencies in neural networks.
• This approach advances the understanding of neural coding and network dynamics.