©
Shutterstock
Wellcome Early Career Research Fellow Dr Andrea Luppi outlines his new paper, which found that parts of the brain work in competition, as well as co-operation, and how this can have implications for the development of digital brain twins and artificial intelligence, which could help advance science and healthcare. The new paper has been published in Nature Neuroscience, was co-authored by an international team including Professor Morten Kringelbach in Oxford and Professor Gustavo Deco.
What is your new study about?
The brain is often described as a highly cooperative system, with specialised regions working together to support cognition and behaviour. However, everyday experience from focusing attention also reveals that brain systems must compete for limited resources.
Our new study shows that this balance between cooperation and competition is a fundamental principle of how the brains of humans and other animals function, which can translate to more brain-like artificial intelligence. Computer models of the brain that are based on this principle are also more faithful to how real brains work, providing a key step towards precision medicine.
In our large comparative study across humans, macaque monkeys, and mice, our international team of researchers used advanced computer modelling to reveal that the most realistic models of brain activity require not only cooperative interactions within specialised brain circuits, but also widespread, long-range competitive interactions between them.
What did you find?
We compared two types of computational brain models: one in which all interactions between brain regions were cooperative, and another in which regions could either excite or suppress each other’s activity. Across all three species, the models that included competitive interactions consistently outperformed cooperative-only models.
Too much cooperation can push the brain into overly synchronised states that are rarely seen in real brains. Competitive interactions act as a stabilising force, preventing runaway activity and allowing different brain systems to take turns in shaping global brain dynamics.
Too much cooperation can push the brain into overly synchronised states that are rarely seen in real brains. Competitive interactions act as a stabilising force, preventing runaway activity and allowing different brain systems to take turns in shaping global brain dynamics.
Remarkably, models with competitive interactions also generated brain activity patterns that closely resembled those associated with real cognitive processes. Using a large-scale analysis of more than 14,000 neuroimaging studies, the researchers showed that spontaneous activity in the competitive models more faithfully reflected known cognitive circuits, such as those involved in attention, memory, and higher-order cognition.
This suggests that competition is crucial for enabling the brain to flexibly activate appropriate combinations of regions: a hallmark of intelligent behaviour.
What is a digital twin brain and how does this study help in their development?
A key advance of the study is its relevance for the development of “digital twins” of the brain: personalised computational models that reproduce the unique activity patterns of an individual’s brain.
The researchers fitted their models to individual human participants using each person’s own structural and functional brain data. Models that included competitive interactions were not only more accurate overall, but also more individual-specific, better capturing what distinguishes one person’s brain from another’s.
This brings us a significant step closer to realistic digital brain twins. Such models could eventually be used to simulate how an individual brain responds to stimulation, medication, or disease, before interventions are applied in the real world. For example, in the future it may become possible to use a computational model to try and predict the effects of a surgical intervention.
What does this study say about other species’ brains?
This has major implications for translational neuroscience. Animal models are routinely used to test treatments before human trials, yet differences between species often limit how well results translate. A modelling framework that works across species provides a powerful bridge between basic research and clinical application.
Importantly, the same cooperative–competitive architecture was found consistently across humans, macaques, and mice. This cross-species convergence suggests that the principles uncovered are fundamental features of mammalian brain organisation, rather than artefacts of a particular dataset or method.
This has major implications for translational neuroscience. Animal models are routinely used to test treatments before human trials, yet differences between species often limit how well results translate. A modelling framework that works across species provides a powerful bridge between basic research and clinical application.
A new foundation for VIRTUAL brain models
The fact that the same findings hold across humans and key animal models suggests that they may reflect fundamental principles of how intelligent systems work. Indeed, the study also shows that in models, cooperative–competitive brain networks have superior capacity to perform a memory task when used as the architecture for a brain-inspired artificial neural network. Networks with competition were better able to process and integrate information, echoing growing evidence that the brain’s balance of cooperation and competition is essential for intelligent computation, and this could apply not only to real brains and model brains, but also systems that aim to replicate the human brains work, like AI. By furthering the link between artificial and biological intelligence, this study advances our mechanistic understanding of how brain network architecture supports information processing.
We concluded that this advance is not only a step forward for neuroscience, but also a critical milestone for the future of personalised brain modelling, digital twins, and translational medicine: potential applications range from understanding disease mechanisms to designing and testing targeted interventions in a virtual environment before they reach patients.
A similar version of this blog written by Dr Andrea Luppi was published on The Conversation.