Inferring context-dependent computations through linear approximations of prefrontal cortex dynamics

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• The prefrontal cortex (PFC) exhibits complex neural dynamics crucial for cognitive functions.
• Understanding how these dynamics arise and support neural computations remains a significant challenge.

Purpose of the Study:

• To infer the underlying mechanisms of context-dependent sensory integration in the PFC.
• To link neural population activity with computational functions through dynamical modeling.

Main Methods:

• Fitting dynamical models to population responses recorded from behaving monkeys.
• Utilizing linear dynamics models driven by external inputs to capture PFC activity.
• Comparing models with context-dependent recurrent dynamics versus contextual input modulation.

Main Results:

• Linear dynamics models accurately captured PFC responses across different contexts.
• Two equally performing mechanisms were identified: transient input amplification and subtle contextual input modulation.
• Both models revealed previously unapparent properties of inputs and recurrent dynamics.

Conclusions:

• Dynamical modeling provides a quantitative framework to understand complex cortical activity.
• The findings offer insights into attentional effects in sensory areas supporting flexible PFC function.
• This approach bridges the gap between neural population dynamics and cognitive computation.