Maintenance of delay-period activity in working memory task is modulated by local network structure

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Understanding the relationship between neural network structure and function is crucial in neuroscience.
• The topological structure of microcircuits in working memory (WM) networks may vary, but its effect on WM activity is unknown.
• Anatomical data suggests gradients within WM networks.

Purpose of the Study:

• To propose a spiking neural network model that replicates fundamental WM characteristics.
• To investigate the impact of structural heterogeneity on WM activity.
• To elucidate the relationship between neural network structure and function.

Main Methods:

• Developed a spiking neural network model to simulate WM.
• Reproduced experimentally observed receptor expression gradients.
• Analyzed correlations between local network structures (small-worldness, excitation-inhibition balance, cycle structures) and WM duration.

Main Results:

• The model successfully replicated delay-period neural activity in association cortex, not sensory cortex.
• Small-worldness, excitation-inhibition balance, and cycle structures were identified as critical for sustaining WM activity.
• Simulations combining anatomical data showed WM duration relies on local and distributed network interactions.

Conclusions:

• Network structural gradients and the interplay between local and distributed networks are vital for working memory.
• Further measurement of structural circuit gradients in the brain is needed.
• The study provides insights into how neural network architecture supports cognitive functions like working memory.