CALLIOPE · BCOM

Functional brain networks exhibit both cooperative and competitive interactions, yet existing models—assuming purely excitatory long-range coupling—fail to account for the widespread anti-correlations observed in fMRI. Starting from a laminar neural mass frame-work, where each mass comprises distinct slow (alpha-band) and fast (gamma-band) oscillatory pyramidal subpopulations (P1 and P2), we show how laminar-specific long-range excitatory projections across neural mass parcels can give rise to both cooperation and competition via cross-frequency envelope coupling. We demonstrate that homologous connections across parcels (e.g., P1→P1 or P2→P2) induce positive correlations between the infra-slow amplitude fluctuations of alpha band envelopes in each parcel, as well as in the simulated fMRI BOLD signals. Conversely, heterologous connections (P1→P2) induce negative correlations. We tested this mechanism by building personalized whole-brain models for a cohort of 60 subjects in two steps. First, we inferred signed inter-parcel generative effective connectivity directly from resting-state fMRI using regularized maximum-entropy (Ising) models. Then we connected laminar neural masses to simulate BOLD dynamics by implementing positive and negative Ising connections via homologous and heterologous projections, respectively. Ising-derived cooperative/competitive connectivity modeling faithfully reproduced both static and dynamic functional connectivity patterns, as well as gamma power-BOLD correlation and partial alpha power-BOLD anticorrelation–outperforming structurally constrained and cooperative-only variants. This further demonstrates that functional data alone suffices to infer individualized connectivity. Together, these results provide a biologically grounded mechanistic model on how long-range excitatory circuits and local cross-frequency interactions shape the balance of cooperation and competition in large-scale brain dynamics.