Minimally dependent activity subspaces for working memory and motor preparation in the lateral prefrontal cortex

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• The lateral prefrontal cortex (LPFC) integrates diverse information, including working memory and motor preparation.
• Mechanisms for downstream regions to isolate specific information types from the LPFC network remain unclear.

Purpose of the Study:

• To investigate how the lateral prefrontal cortex segregates working memory and motor preparation information.
• To identify the neural basis for independent information processing within a mixed-selectivity neuronal population.

Main Methods:

• Analysis of neural activity in the lateral prefrontal cortex of non-human primates.
• Dimensionality reduction techniques to identify neural subspaces.
• Modeling using a bump attractor network with divisive normalization.

Main Results:

• Identified two minimally dependent low-dimensional subspaces within the LPFC: one for working memory, one for motor preparation.
• Both subspaces contained target information during delay periods, which was reduced in error trials.
• A single population of mixed-selectivity neurons formed both subspaces, maintaining largely independent information representation.

Conclusions:

• The lateral prefrontal cortex utilizes distinct neural subspaces to independently encode working memory and motor preparation.
• A bump attractor model with divisive normalization can replicate the observed neural data and information segregation.
• Findings offer novel insights into the computational principles governing prefrontal cortex function and information processing.