Perturbing whole-brain models of brain hierarchy: An application for depression following pharmacological treatment

The article explores how the brain represents information and how this representation changes in neuropsychiatric conditions like major depressive disorder (MDD). Traditionally, neural representation was thought to occur at the level of individual neurons or small groups of cells, with classic examples including place cells in the hippocampus and orientation-selective cells in the visual cortex. However, recent advancements in neuroimaging, particularly functional MRI and magnetoencephalography, have enabled the study of distributed brain activity across larger networks. This shift has introduced the concept of the connectome—a map of neural connections—and has shown that the brain exhibits small-world properties and modular organisation, suggesting a balance between local processing and global integration.

This broader perspective challenges the idea of localised representation and suggests that information may be encoded in hierarchical, distributed networks. Measures such as functional connectivity (FC) and structural connectivity (SC) have identified disruptions in network dynamics in conditions such as Alzheimer’s disease, schizophrenia, PTSD and MDD. Despite their utility, these measures remain descriptive and do not explain how structure produces function. Whole-brain models address this limitation by simulating the interaction of brain regions based on anatomical connectivity and local dynamics. These models are personalised using individual neuroimaging data and can simulate transitions between different brain states, offering insight into causality and adaptability in neural systems.

The authors incorporate the thermodynamics of mind framework into these models, allowing them to estimate the hierarchy of causal influences by examining asymmetries in information flow. Hierarchical systems break a physical principle known as detailed balance, producing neural signals that are irreversible in time. This irreversibility serves as a marker of functional hierarchy and is quantified using temporal asymmetry between time-shifted signals. These features are integrated into generative effective connectivity (GEC) models, which simulate directional and hierarchical dynamics. The authors also apply dynamic sensitivity analysis—a technique for assessing how brain states respond to in-silico perturbations, such as simulated stimulations—to study transitions between depressive and healthier brain states.