Disentangling the Roles of Distinct Cell Classes with Cell-Type Dynamical Systems

Area of Science:

• Computational Neuroscience
• Systems Neuroscience
• Neuroscience

Background:

• Latent dynamical systems model neural population activity but often overlook cell-type diversity.
• Ignoring cell types limits understanding of functional roles and predicting perturbation effects.

Purpose of the Study:

• Introduce the cell-type dynamical systems (CTDS) model to account for distinct neural cell classes.
• Enhance models of neural population dynamics for greater biological interpretability.

Main Methods:

• Developed CTDS, extending latent linear dynamical systems with cell-type-specific latent variables and biologically inspired constraints.
• Applied CTDS to neural recordings from rat frontal orienting fields and anterior dorsal striatum during a decision-making task.
• Validated CTDS by comparing its predictive accuracy and perturbation analysis against standard linear dynamical systems (LDS).

Main Results:

• CTDS demonstrated higher accuracy than standard LDS in analyzing neural recordings.
• Neural choice information was decodable from both excitatory and inhibitory latent variables.
• CTDS accurately replicated in-silico optogenetic perturbation effects, unlike standard LDS.
• The model revealed cell-specific latent dynamics underlying perturbation effects.

Conclusions:

• CTDS provides a more accurate and biologically interpretable framework for neural population dynamics.
• The model enhances the understanding of how distinct cell types contribute to brain function and behavior.
• CTDS can aid in identifying unknown neuronal cell types in electrophysiological recordings.