The brain's anatomical connectivity is relatively stable over short time scales, yet it supports a rich and continuously changing repertoire of functional states underlying diverse cognitive processes and behaviors. How fixed brain structure can both constrain and support this functional flexibility is a longstanding question in systems neuroscience. Previous studies have largely focused on regional structure-function coupling. They therefore cannot directly determine how strongly an instantaneous whole-brain activity state is constrained by global structural topology, or whether that constraint changes systematically as brain states reconfigure.
To address this question, Tianzi Jiang's team at the Institute of Automation, Chinese Academy of Sciences, in collaboration with Yu Zhang's group at Shanghai Jiao Tong University, introduced "structural connectome embeddedness." The concept proposes that whole-brain functional activity does not vary independently of brain structure but evolves within a functional state space shaped by anatomical connections.
To quantify this property, the research team developed connectome-driven embedding (CoDE). CoDE represents the structural connectome as a weighted network and measures the topological smoothness of an instantaneous brain activity pattern over that network. To control for spatial autocorrelation in brain activity patterns, the method generates a null distribution from 10,000 spin-permuted surrogate maps that preserve spatial continuity and converts the observed topological smoothness into a bias-corrected Z score. CoDE therefore provides a time-resolved measure of structural connectome embeddedness that is corrected for spatial smoothness bias; a higher value indicates that a brain activity pattern is more strongly embedded in structural connectome topology.
To determine whether CoDE captures an organizing dimension rather than merely a descriptive feature, the researchers adopted a stepwise validation strategy. They combined a biophysically grounded dynamic mean-field model, 18 established macroscale brain maps, and multimodal imaging and behavioral data from the Human Connectome Project S1200 release. This design allowed the team to move from mechanistic plausibility and broad generality to the functional and behavioral relevance of embeddedness.
The researchers first asked whether embeddedness could arise from neural dynamics constrained by anatomical connections. In the whole-brain model, stronger global structural coupling produced higher average CoDE and a larger proportion of significantly embedded simulated activity patterns within a physiologically stable parameter range. CoDE also varied consistently with firing stability and with the similarity between simulated and empirical functional organization. These results supported the mechanistic plausibility of embeddedness. A complementary test then showed that 17 of 18 brain maps spanning functional, cognitive, electrophysiological and metabolic domains exhibited significant embeddedness, indicating that the phenomenon extends across diverse patterns of macroscale brain organization.
Having established mechanistic plausibility and broad generality, the researchers turned to moment-to-moment brain dynamics in individual participants. Using multimodal HCP imaging data, they found that structural connectome embeddedness was widespread during rest: in 95.93% of participants, more than 85% of time frames reached significant embeddedness. To determine whether this widespread embeddedness had functional significance, the team next examined how different time frames contributed to stable functional connectivity. At the same sampling proportion, connectivity reconstructed from high-CoDE frames more closely resembled the connectivity calculated from the complete scan than did connectivity reconstructed from low-CoDE frames. This advantage was statistically significant across sampling proportions from 1% to 99%. Highly embedded moments may therefore act as "dynamic anchors", maintaining relatively stable functional relationships between brain regions amid continuously changing spontaneous activity. Low-CoDE frames, by contrast, more often corresponded to high-entropy transition points during brain-state reconfiguration. Together, these findings link embeddedness to the balance between stable functional coordination and dynamic state reconfiguration.
Having demonstrated the moment-to-moment functional relevance of embeddedness, the researchers next asked whether its variability also captured meaningful differences between individuals. A multivariate analysis related the temporal variability of CoDE across seven resting-state functional networks to 109 behavioral phenotypes. The resulting pattern corresponded to a behavioral dimension associated with well-being and psychopathology, extending the significance of embeddedness from instantaneous functional organization to individual behavioral variation.
Finally, the researchers tested whether the same organizing dimension extended beyond spontaneous activity to externally driven cognition. Structural connectome embeddedness was similarly widespread in task-evoked brain activity, and its interindividual variability showed a degree of genetic influence. More importantly, CoDE was consistently higher under high- than low-demand conditions in working memory, language and relational reasoning tasks. This pattern suggests that complex cognitive processing may rely more heavily on long-range structural connections to support broader coordination across distributed brain regions.
Overall, the study shows that state-to-state changes in the relationship between functional activity and its underlying structural topology are not incidental fluctuations around a static structure-function relationship. Instead, they may carry systematic information about macroscale functional brain organization. Structural connectome embeddedness provides a quantifiable, state-level description of this variation, allowing the apparent tension between functional diversity and structural stability to be viewed as a potentially informative continuum of large-scale brain organization. This framework therefore offers a new perspective on how a relatively stable structural connectome supports the flexible organization of brain activity.
Weiyang Shi from the Institute of Automation, Chinese Academy of Sciences, is the first author. Tianzi Jiang from the Institute of Automation and Yu Zhang from Shanghai Jiao Tong University are the co-corresponding authors. The study was supported by the Science and Technology Innovation 2030 - Brain Science and Brain-Inspired Intelligence Project, the National Natural Science Foundation of China, and the China Postdoctoral Science Foundation.
Science Bulletin
Data/statistical analysis