Gaussian Process Inference Reveals Non-Separability of Position and Velocity Tuning in Grid Cells

Area of Science:

• Neuroscience
• Computational Neuroscience
• Spatial Cognition

Background:

• Grid cells in the medial entorhinal cortex (MEC) are crucial for spatial navigation.
• These cells respond to various factors, including position, speed, and head direction.
• Understanding the combined (conjunctive) coding of these variables is limited.

Purpose of the Study:

• To investigate the conjunctive coding of position and velocity in MEC grid cells.
• To develop methods for analyzing high-dimensional neural tuning data.
• To determine if grid cell tuning is separable or interactive across position and velocity.

Main Methods:

• Analysis of neural recordings from freely foraging rats.
• Construction of four-dimensional (4D) tuning curves across 2D position and 2D velocity.
• Application of Gaussian Process (GP) methods to estimate firing rates in a large behavioral space.

Main Results:

• Some grid cells exhibit significant non-separability in their position and velocity tuning.
• Gaussian Process modeling revealed interactions not apparent in 2D analyses.
• A data coverage threshold was identified as necessary for observing non-separability.

Conclusions:

• Grid cell coding is not always separable across position and velocity.
• Advanced computational methods like GPs are essential for uncovering complex neural representations.
• This study advances our understanding of how the brain represents space and movement.