Simulating the Living Brain
Simulating the Living Brain
How connectomics and AI are creating digital twins of biological brains, beginning with the remarkable success of the fruit fly.
“ A groundbreaking computational model of the fruit fly brain demonstrates how neural wiring alone can predict behaviour, offering a glimpse into the future of digital brain simulation. ”
|
Nevalor Post
An Editorial Feature · Nevalor Publishers
|
Scientists Build a Digital Twin of the Fruit Fly Brain: A Major Leap Towards Simulating Intelligence
From mapping every neural connection to accurately predicting behaviour, researchers are transforming the dream of simulating a living brain into a scientific reality.
Introduction
For decades, one of neuroscience's greatest ambitions has been to answer a seemingly impossible question: Can a computer truly simulate a living brain?
What once belonged to the realm of science fiction is now becoming an engineering challenge driven by extraordinary advances in neuroscience, artificial intelligence, and computational modelling. Researchers are no longer asking whether it is theoretically possible—they are building increasingly accurate digital replicas of biological nervous systems.
The latest milestone comes from an international team of scientists who have developed a comprehensive computational model of the adult fruit fly (Drosophila melanogaster) brain. Using one of the world's most detailed neural wiring maps, they demonstrated that a computer can accurately predict how sensory information is transformed into behaviour.
Although the fruit fly's brain is tiny compared with that of a human, this achievement represents one of the clearest demonstrations yet that understanding the brain's wiring may be sufficient to predict its function.
Why the Fruit Fly?
At first glance, the humble fruit fly may seem an unlikely candidate for one of neuroscience's most ambitious projects. Yet Drosophila melanogaster has become one of the most valuable organisms in biological research.
Its nervous system is complex enough to perform sophisticated behaviours—including learning, memory, navigation, feeding, courtship, and decision-making—while remaining vastly simpler than the human brain.
An adult fruit fly possesses roughly 200,000 neurons, compared with approximately 86 billion neurons in the human brain. This smaller scale makes it possible to reconstruct every neuron and every connection with remarkable precision. Scientists have long viewed the fruit fly as the ideal stepping stone towards understanding larger and more complex brains.
The FlyWire Connectome: A Complete Wiring Diagram
Every brain simulation begins with a fundamental requirement: knowing how its neurons are connected. This complete map of neural wiring is known as a connectome. The newest computational model is based on the FlyWire connectome, one of the most detailed biological datasets ever produced.
FlyWire maps more than:
- 125,000 reconstructed neurons
- Over 50 million synaptic connections
- The identities of numerous neurotransmitters used throughout the brain
Unlike earlier connectomes, FlyWire provides sufficient detail to follow information from sensory organs through intermediate processing centres and finally to the neurons responsible for movement. This wiring diagram serves as the blueprint for the brain's digital twin.
Building a Functional Digital Brain
Simply possessing a wiring diagram does not automatically produce a functioning brain simulation. Researchers constructed a computational framework using a leaky integrate-and-fire neural model, one of the most widely used mathematical descriptions of neuronal activity.
In this approach, each neuron continuously accumulates incoming electrical signals. Once the accumulated activity reaches a threshold, the neuron generates an electrical impulse before resetting its internal state. Despite its relative simplicity, this model successfully reproduced complex sensorimotor computations across the entire fly brain.
Remarkably, researchers relied primarily on two pieces of biological information:
- The physical connectivity between neurons
- The neurotransmitter identity associated with each connection
No extensive behavioural training was required. The results suggest that anatomical connectivity alone contains far more functional information than previously believed.
Predicting Behaviour from Brain Wiring
The greatest achievement of the model lies in its ability to accurately predict behaviour directly from neural architecture. Researchers simulated how different sensory stimuli propagated through the fly's neural circuits and identified the motor responses that should follow.
The model successfully reconstructed pathways involved in several essential behaviours, particularly:
- Feeding
- Grooming
- Taste processing
Even more impressively, it identified the specific neurons responsible for these behavioural outputs before experimental validation. This predictive capability represents a significant advance towards understanding how brains transform sensation into action.
Sweet, Bitter, and Shared Neural Circuits
One particularly fascinating discovery involved how the fruit fly processes taste. The simulations revealed that pathways responding to sugar and water converge upon many of the same neural circuits before generating feeding behaviour.
This suggests that the fly utilises a shared decision-making network for multiple rewarding stimuli. By contrast, aversive tastes, such as bitter compounds, remained largely isolated within separate neural pathways. Such segregation likely enables rapid avoidance responses without interference from circuits promoting feeding.
These findings provide new insight into how brains balance competing behavioural drives.
Experimental Validation
Computer simulations are valuable only if they accurately reflect biological reality. To verify the model, scientists conducted extensive laboratory experiments using optogenetics, a technique that allows researchers to activate or silence specific neurons using light.
Behavioural experiments confirmed the model's predictions with impressive accuracy. Perhaps most notably, the simulations identified previously unknown inhibitory functions within certain taste-sensitive neurons. Subsequent experiments confirmed these unexpected roles, demonstrating that the computational framework could generate genuinely new scientific discoveries rather than merely reproduce existing knowledge.
This marks an important shift in neuroscience—from using experiments solely to discover brain function towards allowing computational models to predict functions that experiments then verify.
Beyond the Fruit Fly: The Rise of Digital Brain Twins
The fruit fly model forms part of a much broader scientific movement towards creating digital replicas of biological nervous systems. These "digital twins" combine detailed anatomical maps with advanced computational models capable of reproducing neural activity. Their long-term purpose extends beyond insects. Researchers hope these systems will eventually help explain how mammalian—and ultimately human—brains process information.
The Evolution of Connectomics
Modern brain simulation owes its existence to decades of progress in connectomics, the science of mapping neural connections.
OpenWorm: The First Complete Nervous System
One of the earliest milestones was the OpenWorm project. Its objective was to digitally reconstruct the nervous system of Caenorhabditis elegans, a microscopic worm possessing only 302 neurons. Although tiny, this organism demonstrated that complete nervous system simulations were technically achievable.
The Hemibrain Project
As imaging technologies improved, researchers tackled increasingly larger brains. The hemibrain project produced the first dense reconstruction of a substantial portion of the fruit fly brain, cataloguing thousands of neurons and neural pathways.
It provided the foundation upon which FlyWire later expanded.
FlyWire
FlyWire represents one of neuroscience's most ambitious collaborative achievements. Rather than reconstructing only a section of the brain, researchers assembled an almost complete neuronal wiring diagram of the adult fruit fly.
This enormous dataset enables scientists to trace information continuously from sensory receptors to behavioural outputs.
Human Connectomics: The H01 Project
Mapping the human brain remains vastly more difficult. One recent milestone involved reconstructing a single cubic millimetre of the human cerebral cortex. Despite representing only a microscopic fraction of the brain, this tiny volume generated nearly one petabyte of imaging data.
Considering that the human brain contains approximately 86 billion neurons, a complete connectome remains an immense scientific challenge.
New Imaging Technologies
Electron microscopy remains the gold standard for reconstructing neural circuits due to its exceptional resolution. However, newer techniques such as LICONN (Light-Microscopy-Based Connectomics) aim to accelerate mapping while simultaneously identifying molecular characteristics of neurons.
These innovations may dramatically increase the speed of future connectome construction.
Why a Connectome Alone Is Not Enough
Although wiring diagrams provide essential structural information, many neuroscientists caution that anatomy represents only part of the story. A functioning brain depends upon numerous additional biological processes.
Among them are:
- Neurotransmitters released at synapses
- Receptor types on receiving neurons
- Hormonal and neuromodulatory signals
- Electrical gap junctions
- Changes caused by behavioural state
- Cellular biochemistry
- Short-term and long-term plasticity
- Previous neural activity
Together, these factors determine how identical wiring can produce different behaviours under different circumstances. Understanding the brain therefore requires more than simply identifying connections. It requires modelling their dynamic interactions over time.
Forecasting Neural Activity with Artificial Intelligence
Artificial intelligence has become indispensable in modern neuroscience. Beyond reconstructing connectomes, AI now predicts neural activity itself.
ZAPBench
The ZAPBench benchmark evaluates algorithms capable of forecasting activity across an entire larval zebrafish brain. Containing over 70,000 neurons, this system enables researchers to compare different machine learning approaches.
Current evidence suggests that models analysing neural activity as three-dimensional "brain movies" outperform conventional time-series techniques.
Neural Foundation Models
Inspired by large language models, scientists are developing foundation models for neural activity. Rather than learning language, these systems learn the statistical rules governing brain dynamics. Recent work involving the mouse visual cortex has demonstrated exceptional accuracy when predicting responses to unfamiliar visual stimuli. These results suggest AI can learn the complex nonlinear computations performed by biological brains.
Reduced-Order Modelling
Simulating every molecule inside every neuron remains computationally impossible. To overcome this limitation, projects such as OpenWorm employ reduced-order models, including recurrent neural networks that replicate neural behaviour without simulating every biological detail.
These simplified systems capture essential input-output relationships while remaining computationally practical.
Artificial Intelligence and Neuroscience: A Two-Way Partnership
The relationship between neuroscience and AI has become increasingly intertwined. Early artificial neural networks were inspired by biological brains. Today, AI helps neuroscientists reconstruct connectomes by automatically identifying neurons within enormous electron microscopy datasets.
At the same time, neuroscientists increasingly use biological connectomes to design better AI systems. In the fruit fly, connectome-constrained neural networks have successfully reproduced aspects of visual motion processing with far fewer adjustable parameters than traditional machine learning models.
This reciprocal exchange continues to benefit both fields.
The Future of Whole-Brain Simulation
The successful simulation of the adult fruit fly brain represents far more than an isolated technical achievement. It demonstrates that accurate behavioural prediction can emerge from biological connectivity combined with relatively simple computational principles.
As connectomic datasets become larger and computational methods more sophisticated, scientists move steadily closer to constructing functional digital twins of increasingly complex nervous systems.
Human brain simulation remains many years—perhaps decades—away. Nevertheless, every advance in insects, worms, zebrafish, and mice provides valuable insights into the universal principles governing neural computation.
Conclusion
The development of a comprehensive computational model of the adult Drosophila brain marks a defining moment in neuroscience. By combining the FlyWire connectome with biologically informed neural modelling, researchers have shown that the brain's physical wiring and neurotransmitter identities alone can accurately predict behaviour.
Validated through optogenetic and behavioural experiments, the model not only reproduced known sensorimotor pathways but also uncovered previously unknown neural functions, highlighting its value as a powerful scientific discovery tool.
While a complete digital simulation of the human brain remains a distant ambition, these advances demonstrate that increasingly faithful virtual brains are becoming a reality. Each new connectome, computational model, and AI-driven prediction brings neuroscience closer to understanding one of nature's most intricate systems—not by examining isolated neurons, but by revealing how billions of connections work together to produce perception, decision-making, and behaviour.
The fruit fly's digital twin is therefore much more than an insect simulation. It is a glimpse into the future of neuroscience, where virtual brains may one day allow researchers to test hypotheses, investigate neurological disorders, and explore the mechanisms of intelligence in ways that were once unimaginable.
Editorial Note
References and further study
Azevedo, Frederico A. C., Suzana Carvalho, Lea T. Grinberg, José M. Farfel, Renata E. Ferretti, Rafael E. P. Leite, Wilson Jacob Filho, Roberto Lent, and Suzana Herculano-Houzel. 2009. “Equal Numbers of Neuronal and Nonneuronal Cells Make the Human Brain an Isometrically Scaled-Up Primate Brain.” Journal of Comparative Neurology 513 (5): 532–541.
Dorkenwald, Sven, Louis K. Scheffer, Igor Pisarev, David B. Turner, Philipp Schlegel, and FlyWire Consortium. 2024. “The FlyWire Whole-Brain Connectome of Adult Drosophila.” Nature. DOI: https://doi.org/10.1038/s41586-024-07558-y
OpenWorm Foundation. OpenWorm Project. Accessed August 7, 2026. https://openworm.org
Shapson-Coe, Alexander, MichaÅ‚ Januszewski, Daniel Berger, et al. 2024. “A Petascale Reconstruction of Human Cerebral Cortex.” Nature. DOI: https://doi.org/10.1038/s41586-024-07663-y
Tschopp, Fabian D., and colleagues. 2025. “A Whole-Brain Computational Model of the Adult Drosophila Brain Predicts Sensorimotor Transformations from the Connectome.” Nature.
Varshney, Lav R., Beth L. Chen, Eric Paniagua, David H. Hall, and Dmitri B. Chklovskii. 2011. “Structural Properties of the Caenorhabditis elegans Neuronal Network.” PLoS Computational Biology 7 (2): e1001066.
Xu, C. Shan, Kenneth J. Hayworth, Zhiyuan Lu, and colleagues. 2021. “Enhanced FIB-SEM Systems for Large-Volume 3D Connectomics.” Nature Methods 18 (1): 21–26.
Comments