Networking in primary motor cortex predicts arm-reaching speed variation

Abstract

Neuronal ensemble activity of the primary motor cortex (M1) produces intricate patterns associated with the speed of arm movements. Unlike directional tuning properties of single neurons, collective changes of the population of neurons are known to characterize speed. However, such a collective activity may or may not contain neurons directly coding speed information. For example, trial-by-trial variability of the speed profile of arm movements is predominantly predicted by preparatory activity, but without detailed aspects of the maximum speed and its latency. To address these issues, we investigated kinematics-related latent components of neuronal populations and their interactive patterns. Kinematics-related latent components were estimated by canonical correlation analysis (CCA). Then, we analyzed connectivity between latent variables using Pearson correlation coefficients, from which we formed a network of latent components. We further quantified the characteristics of the network by a clustering coefficient based on the graph theory, which reflects the degree of clustered networks and node density. Dataset for this study was acquired from M1 of a rhesus macaque that performed a two-dimensional center-out reaching task, which is available at a public database on the collaborative research in computational neuroscience (CRCNS). To analyze the variability of the speed profile, we separated individual profiles into three disjoint groups having early- (≤ 25 %), mid- (> 25 % or < 75 %) and late- (≥ 75 %) latency of the maximum speed regardless movement direction. Plus, the trials of the speed magnitude were also separated in the same manner into three groups. We found that the clustering coefficient increased as it took longer to reach the maximum speed in arm movements. We also found that the clustering coefficient increased as the maximum speed increased. Our results suggest that the property of networking among latent components of peri-movement M1 population activities may elucidate the variability of the speed profile of point-to-point arm movements with an identical distance.