Amplitude-Conditioned Brain Dynamics Reveal Widespread Negative Modulation and Targeted Positive Modulation of...

Functional connectivity is traditionally characterized through pairwise interactions between brain regions. However, growing evidence suggests that neural communication is also shaped by higher-order interactions in which the activity of a third region modulates the coupling between others. Here, we introduce a whole-brain framework for mapping resting-state nonlinear conditional connectivity from functional MRI. For each cortical region, we quantified how spontaneous amplitude fluctuations modulate the functional connectivity of all other region pairs. This procedure generated a modulation spectrum for every connection, describing connectivity changes across multiple amplitude states of the modulator. From these spectra, we derived four interpretable features capturing modulatory strength, polarity, volatility, and nonlinear complexity. The framework was applied to resting-state data from the Human Connectome Project and evaluated using stationary surrogate models, spatial topography analyses, test-retest reliability assessments, and behavioral prediction. Surrogate analyses demonstrated that the observed modulation spectra could not be explained by static covariance structure or spectral properties alone, indicating genuine state-dependent dynamic interactions. Topographic mapping reveals a profound functional asymmetry: the macroscopic connectome is dominated by widespread negative modulatory influences, with positive modulation restricted to sparse assortative pathways. Furthermore, by evaluating the effect of temporal integration, we show that spatial and temporal aggregation robustly stabilizes these dynamic fluctuations into reliable, individual-specific traits. Finally, employing regularized predictive modeling, we demonstrate that these dynamic modulatory features can provide complementary predictive value beyond conventional static FC, particularly for capturing state-dependent affective traits such as internalizing behavior. Together, this framework unveils a structured, asymmetric, and behaviorally relevant higher-order modulatory architecture.