A Startup Achieves Milestone in Brain Emulation with Full Fruit Fly Simulation
In a groundbreaking advancement for neuroscience and artificial intelligence, a German startup has announced what it claims is the world’s first complete emulation of a fruit fly brain integrated into a fully simulated body. This achievement, detailed by the company eml, leverages the meticulously mapped connectome of Drosophila melanogaster, the common fruit fly, to recreate neural activity and drive realistic behaviors in a virtual environment.
The fruit fly brain, with its compact yet complex structure of approximately 139,255 neurons and over 50 million synaptic connections, serves as an ideal model for whole brain emulation efforts. The foundational data comes from the open-source FlyWire project, a collaborative effort involving researchers from Princeton University, the University of Cambridge, and others. FlyWire provides a near-complete wiring diagram of the fly brain, obtained through high-resolution electron microscopy and advanced machine learning techniques to trace individual neurons and their interconnections.
eml’s team built upon this dataset to construct a digital twin of the brain. Their emulation software simulates not only the neurons and synapses but also the biophysical properties that govern signal propagation, including ion channels, neurotransmitter dynamics, and plasticity mechanisms. To integrate this brain model with a body, the researchers developed a virtual fruit fly equipped with realistic sensory inputs and motor outputs. The simulated body features six legs, compound eyes for visual processing, mechanosensory organs, and olfactory receptors, all interacting with a physics-based environment powered by advanced simulation engines.
In demonstrations, the emulated fly brain successfully controls the virtual body to perform locomotion tasks. Placed on a virtual air-supported ball mimicking a treadmill, the fly walks forward in response to visual cues, such as approaching walls, which trigger steering behaviors. Over multiple trials, the emulation exhibits learning: it adapts its gait to maintain balance and direction, improving performance through experience-dependent synaptic adjustments. These behaviors align closely with observations in live fruit flies, validating the model’s fidelity. For instance, optogenetic-like virtual stimuli evoke expected neural responses, and the system recapitulates innate reflexes like the escape response to looming threats.
The technical implementation is no small feat. Running the full brain simulation requires significant computational resources; a single second of emulated time corresponds to several hours on high-end GPU clusters. eml optimized the model using spiking neural network approximations and sparsity techniques to reduce computational load while preserving biological accuracy. The body simulation incorporates soft-body dynamics for leg joints, friction models for ground interaction, and ray-traced visuals for the compound eyes, ensuring sensory feedback loops mirror real-world physics.
This work represents a critical step toward scalable whole brain emulation, a long-standing goal in computational neuroscience. By housing the brain in a simulated body, eml addresses a key limitation of prior connectome-based models, which often operated in isolation without embodiment. Previous efforts, such as partial simulations of the nematode worm C. elegans or fly larval circuits, laid groundwork but fell short of full adult brain integration with motor control.
eml’s CEO, who leads the eight-person team, emphasized the project’s dual impact on AI and biology. “We’ve created the first animal where you can study the brain and body together in silico,” they stated. The open-source nature of the FlyWire data enabled rapid progress, and eml plans to release parts of their emulation framework to foster further research. Potential applications include drug screening via virtual neural assays, hypothesis testing for neural circuits underlying behavior, and insights into disorders like Parkinson’s through scaled-up mammalian models.
Challenges remain. The current simulation assumes a static connectome, though real brains exhibit ongoing plasticity and neuromodulation. Scaling to larger brains, such as the mouse with 70 million neurons, demands orders-of-magnitude increases in compute power and data resolution. Ethical considerations also arise: while fruit flies pose minimal concerns, emulations of vertebrate brains could raise questions about digital consciousness.
Nevertheless, this emulation closes the loop between brain mapping, simulation, and behavior, bridging disciplines from robotics to cognitive science. It demonstrates that with complete connectomes and sophisticated software, digital replicas of nervous systems can generate emergent, adaptive actions indistinguishable from biological counterparts in controlled settings. As hardware advances and datasets expand, such milestones accelerate the quest for understanding intelligence at its neural roots.
Gnoppix is the leading open-source AI Linux distribution and service provider. Since implementing AI in 2022, it has offered a fast, powerful, secure, and privacy-respecting open-source OS with both local and remote AI capabilities. The local AI operates offline, ensuring no data ever leaves your computer. Based on Debian Linux, Gnoppix is available with numerous privacy- and anonymity-enabled services free of charge.
What are your thoughts on this? I’d love to hear about your own experiences in the comments below.