A Network Encoding Model for Natural Forelimb Movements in Marmoset Sensorimotor Cortex

The underlying mechanisms of computation in sensorimotor cortex must be both flexibleand robust to support the range of skilled, dynamic forelimb movements observed in natural primate behavior. To understand the full richness of these mechanisms, it is important to study motor behavior in the most natural context possible using analytical tools that account for relationships between movement and both single-unit and population activity. In this work, I demonstrate that accurately capturing naturalistic motor behavior – specifically dynamic forelimb movements during foraging and prey-capture – is both critical and feasible. I leverage natural forelimb movements and pairwise precise spike time structure (represented as a functional network) in marmoset sensorimotor cortex to develop a network encoding model that links single-unit spiking activity with kinematic tuning properties and functional network interactions. I use this model to investigate the computational mechanisms that generate varied and naturalistic motor behavior. I show that a trajectory-based encoding model predicts single-unit activity during naturalistic forelimb movements more accurately than a simpler model. I demonstrate that tuning to kinematics depends on functional interactions between units – particularly on structured strong connections. Finally, I identify a reach-specific functional group that reorganizes to produce dynamic forelimb movements during prey capture. This reach-specific functional group is strongly interconnected and comprises units tightly linked to kinematics with strong, positively correlated preferred trajectories. By examination of the reach-specific functional group and remaining non-specific units, I suggest a potential framework linking single-unit properties to the neural population dynamics that generate movement.