Linking neural manifolds to circuit structure in recurrent networks

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Neural Circuit Dynamics

Background:

• Dimensionality reduction is key to understanding neural activity, examining single-neuron properties and population dynamics.
• The relationship between neural circuit structure and these activity aspects remains unclear.

Purpose of the Study:

• To link neural circuit structure with single-neuron functional properties and emergent low-dimensional population dynamics.
• To investigate how distinct circuit architectures can lead to similar dynamic patterns.
• To establish criteria for evaluating neural network models against empirical data.

Main Methods:

• Development of spiking recurrent network models.
• Analysis of the relationship between circuit topology, single-neuron properties, and population-level dynamics.
• Comparison of model predictions with observed neural activity.

Main Results:

• Topologically different circuit structures can generate equivalent low-dimensional dynamics.
• Circuit structure imposes constraints on both population activity dynamics and single-neuron functional properties.
• Identified simple criteria for model-data comparison.

Conclusions:

• The study connects classical circuit models with the concept of neural manifolds.
• Provides a framework for creating more accurate models of neural population dynamics aligned with recordings.
• Highlights the crucial role of circuit structure in shaping neural activity patterns.