A TOPOLOGICAL FINGERPRINT ENCODES MOTOR SKILL AT REST

Spontaneous brain activity at rest exhibits complex patterns of interactions, whose functional role is under debate [1]. A prominent hypothesis is that resting-state connectivity encodes behavioral performance and cognitive abilities [2,3]. In this study, we investigated whether resting-state MEG brain network topology could model individual manual dexterity in 86 healthy subjects of the Human Connectome Project. Functional connectivity was estimated using band-limited power correlations in alpha, low-beta, and high-beta bands, and network topology was characterized through Participation Index (PI) of connector hubs after modular decomposition [4]. A machine learning regression framework with exhaustive feature selection and repeated stratified holdout validation was applied to extract a low-dimensional fingerprint predicting dexterity scores [5]. We identified an optimal fingerprint in the alpha band, consisting of four PI-based parietal hubs, that accurately modelled individual manual dexterity (r = 0.34). A multi-band model achieved comparable performance but required three times more features. A vulnerability analysis simulating targeted hub disconnections revealed that two hubs (namely RSPL-preCun and RvIPS) were critical, as their removal caused a marked drop in predictive performance. We combined these features to propose a functional “refocusing” mechanism: hubs progressively prune connections external to their modules when dexterity increases, while maintaining an internal representation of dexterity performance. Such inhibition and maintenance are well aligned with the role of the alpha band reported in the literature [6]. These findings suggest that the architecture of interactions at rest, by combining few topological features in the alpha band, encodes stable behavioral traits, such as manual dexterity.