Inferring the intrinsic neural timescales of the human connectome using optimal control theory

The human connectome describes a map of whole-brain structural connectivity [1]. At the macroscale, this network map comprises brain regions—nodes—whose local biological properties and interconnectivity— edges—determine their capacity to communicate with one another. Modeling this complex structure- function relationship remains an outstanding challenge in neuroscience. To study this relationship, we have developed and implemented an approach called Network Control Theory (NCT) [2]. NCT stipulates neural dynamics that evolve on the connectome as a function of (i) nodes’ intrinsic dynamics and (ii) their extrinsic connectivity, and models control inputs that guide those dynamics to transition between empirically measured brain states. In NCT, nodes' intrinsic dynamics are encoded by their internal decay rate, which governs their intrinsic neural timescales (INTs) and can be thought of as their propensity towards self-inhibition; higher self-inhibition yields faster dissipation of perturbations, which equates to faster INTs. Conventionally, internal decay rates are set uniformly across the connectome, yielding nodes that exhibit consistent self-inhibition and, thus, uniform INTs. However, this setup is at odds with our understanding of how neuronal time scales vary across the brain in tandem with diverse spatially patterned properties of neurobiology [3]. Thus, to improve the biological realism of NCT, we developed a data-driven approach to optimizing nodes' internal decay rates. Compared to previous work [4], our approach does not require a priori knowledge of neurobiology. This independence allows us to validate our optimized INTs against known neurobiological correlates, and to fit our model on a per-transition and a per-subject basis. We found that using optimized decay rates yields lower control energy associated with transitions between empirical brain states. Further, we found that optimized decay rates coupled to in vivo empirical measures of INTs as well as ex vivo measures of gene expression and cell-type densities that underpin structure- function coupling [5]. Finally, when applied to single-subject connectomes, our approach significantly improves the out-of-sample prediction of behavior in healthy young adults. In summary, we provide a novel extension to NCT to better model and validate the complex interaction between brain structure and function with subject-level specificity.
References
1. Sporns, O., Tononi, G. & Kötter, R. The Human Connectome: A Structural Description of the Human Brain. PLoS Comp Biol 1, e42 (2005).
2. Parkes, L. et al. A network control theory pipeline for studying the dynamics of the structural connectome. Nat Protoc (2024) doi:10.1038/s41596-024-01023-w.
3. Wolff, A. et al. Intrinsic neural timescales: temporal integration and segregation. Trends in Cognitive Sciences S1364661321002928 (2022) doi:10.1016/j.tics.2021.11.007.
4. Luppi, A. I. et al. Contributions of network structure, chemoarchitecture and diagnostic categories to transitions between cognitive topographies. Nat. Biomed. Eng (2024) doi:10.1038/s41551-024-01242- 2.
5. Zhang, X.-H. et al. The cell-type underpinnings of the human functional cortical connectome. Nat Neurosci (2024) doi:10.1038/s41593-024-01812-2.