3D Fruit Fly on macOS Desktop Powered by Real FlyWire Connectome
Interactive 3D fly experience animated with actual neural connectivity data.
1Real Neural Network Simulation
The project creates a 3D model of a fruit fly brain using 23,210 out of 139,255 neurons from the FlyWire v783 dataset. A 668-neuron circuit operates with leaky-integrate-and-fire (LIF) model updating 1000 times per second. The circuit includes visual neurons like LC4 and LPLC2, Giant Fiber escape command neurons, and various walking/turning motor neurons.
2Biological Behavior Accuracy
The fly mimics neural responses of real fruit flies. For example, it shows escape response in ~4ms when cursor approaches slowly. This is achieved through actual synaptic connections of Giant Fiber neurons. The fly also exhibits biological behaviors like grooming, sleep postures, and speed adjustment based on temperature.
3Technical Specifications and Interaction
The application requires macOS 13+ and Xcode Command Line Tools. Users can interact with mouse, stimulate neurons by clicking brain regions. The fly walks across windows, perches on edges, and responds to window movements. The system operates on a completely transparent overlay without requiring permissions.
4Educational Potential
The project offers unique tool for neuroscience education. Users can interactively observe how real neural circuits function. Particularly, the Giant Fiber escape circuit, LC4/LPLC2 visual neurons and various motor control neurons provide ideal environment for understanding basic nervous system modules.
5Modeling vs Real Data
The project uses FlyWire Codex data dumps. While the connection graph is real, LIF dynamics, neurotransmitter signs and synaptic delays are modeling choices. Data is processed from ~60MB raw dumps using python etl.py command. The system also provides simulation tests and behavior tests.
6Behavior Diversity
The fly exhibits 17 different end-to-end tested behaviors. It shows specific responses when neurons are stimulated: escapes when Giant Fiber fires, grooms when DNg11 is stimulated, turns when DNa01/02 are activated. Body movements feed back to brain: walking rhythm stimulates ascending (proprioceptive) neurons and fast cursor motion triggers wind-sense neurons.
7System Requirements
The project is developed with Swift 5.9+ and open-sourced on GitHub. Installation requires git clone and simple build.sh command. The app is controlled from menu bar: offers pause, show brain window, escape test, display switching and add/remove fly features. System resource usage is optimized.
8Neuroscience Meets Technology
This project pioneers using real neural connectomes in interactive simulations. By combining biological accuracy with software engineering, it presents new paradigm for understanding complex nervous systems. Particularly inspirational for future projects in education, research and neuromorphic computing fields.
Original article: Hacker News — Gerçek Sinek Beyniyle Çalışan macOS Masaüstü 3D Sineği