A new mathematical model endows Digital Twins of the human brain with self-regulation
Oct. 15, 2026
- An international study co-led by IFISC (UIB-CSIC) incorporates a homeostatic plasticity rule into a biophysical whole-brain model, allowing the virtual brain to dynamically adjust its own excitation-inhibition balance.
- Published in Network Neuroscience and developed in collaboration with the Paris Brain Institute (ICM), the work lays the mathematical foundation for simulating complex transitions between states of consciousness such as sleep, wakefulness, and dissociative conditions.
An international team of neuroscientists and computational physicists, including research from the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC, UIB-CSIC), has developed a new whole-brain biophysical model termed Homeostatic Dynamic Mean Field (HDMF). The new model –expanding on the Dynamic Mean Field model– addresses one of the major theoretical challenges in computational neuroscience: how the brain endogenously maintains the stability of global neural activity under external perturbations or dynamic changes?.
In real biological networks, neural circuits continuously adapt their connections, which is the base for learning and to adequate respose to the environment. However, this process is strictly controlled by homeostasis, the self-regulating capacity of living systems to maintain stable internal conditions. In the brain, homeostasis acts like a "thermostat" that prevents two dangerous extremes: hyperexcitability (an uncontrolled surge of electrical activity that can lead to epileptic seizures or tissue damage) and quiescence (an almost complete shutdown of neural activity). Until now, most large-scale computer models relied on externally imposing inhibitory parameters. These required complex prior calculations and often desynchronized or crashed when simulating perturbations such as drugs, metabolic changes, or network fluctuations.
To overcome this limitation, the team embedded a self-tuning inhibitory plasticity rule into the standard Dynamic Mean Field (DMF) framework, a widely used whole-brain model in human computational neuroscience. This mechanism (HDMF) acts like an automatic feedback control system that continuously balances excitatory and inhibitory signals across every brain region in real time. The model not only preserves fitting accuracy when reproducing functional connectivity measured by functional magnetic resonance imaging (fMRI), but also significantly expands the range of brain states that can be simulated on a computer.
Among its main findings, the study shows that the HDMF model can spontaneously generate slow-wave activity characteristic of deep sleep or general anesthesia, a feature previously unattainable within the standard DMF framework. Furthermore, by introducing regional heterogeneity in adaptation capacity across different areas of the brain, the model enabled the simulation of coexisting dynamic regimes, termed chimeric-like states: the simultaneous coexistence of regions exhibiting slow-wave oscillations (typical of sleep) and others displaying wake-like asynchronous activity. This opens new avenues for modeling dissociative states of consciousness for the first time, such as sleepwalking or parasomnias, where both sleep and wake-like neural dynamics coexist in the brain .
"Incorporating homeostatic self-regulation into the equations not only makes the simulations more robust and biophysically plausible, but also vastly expands the range of brain states we can explore within a single unified framework" explains Rubén Herzog, researcher at IFISC (UIB-CSIC), Distinguished Professor at the Department of Psychology of the UIB, and supervisor author of the study.
This theoretical framework published in Network Neuroscience provides the mathematical foundation supporting a recent parallel study published in Nature Biomedical Engineering, where this same HDMF model is deployed to construct personalized "digital twins" from brain scans of nearly 7,000 individuals. Both studies form part of a unified research line aimed at refining biophysical brain models to advance from theoretical neuroscience toward personalized medicine and in silico discovery of patient-tailored treatments.
Figure: HDMF model dynamics driven by local adaptation speed: from slow-wave sleep-like activity (top) and irregular oscillations (middle), to asynchronous wake-like activity (bottom).
Mindlin, I., Coronel-Oliveros, C., Llabrés, M., Sitt, J. D., & Herzog, R. (2026). The impact of homeostatic inhibitory plasticity in a generative biophysical model. Network Neuroscience, 1–65. https://doi.org/10.1162/NETN.a.611
