BRAVLi
Drosophila Whole-Brain Digital Twin · 2017–present
From Blue Brain to Fly Brain
BRAVLi began at the Blue Brain Project (EPFL, 2017–2024) as a framework for managing computational provenance across petabytes of neuroscience data — 4.2 million morphologically detailed neurons, 14 billion synapses. It has since evolved into something more ambitious: a literate codebase for reconstructing and simulating the Drosophila whole-brain connectome.
The Real Story Lives in codev/
This is a literate codebase. The Python files in bravli/ are tangled output from org-mode lessons in codev/. To understand the code, read the lessons. To modify the code, edit the lessons and re-tangle. The lessons are the documentation, the specification, and the source — all in one place.
18 Lessons, Bottom to Top
| Lesson | Topic |
|---|---|
| 00 | Foundations — dataset abstraction, logging |
| 01 | Parcellation — neuropil region tree, FlyWire loader |
| 02 | Composition — cell type counts, neurotransmitter profiles |
| 03 | Factology — structured measurement system |
| 04 | Visualization — 3D rendering via navis |
| 05 | Mushroom body exploration — integration walkthrough |
| 06 | Atlas — brain atlas and neuropil geometry |
| 07 | Research roadmap — plan for fly brain 2026 |
| 08 | Connectivity — synaptic connectivity analysis |
| 09 | Synaptic physiology — synapse models and neurotransmitters |
| 10 | Cell models — LIF and graded cell models |
| 11 | Simulation — LIF simulation engine |
| 12 | Portal — interactive digital twin portal |
| 13 | Mushroom body circuit — microcircuit extraction + sparseness |
| 14 | ISN and learning — ISN regime + three-factor STDP conditioning |
| 15 | Brunel phase diagram — FlyWire regime mapping |
| 16 | Neuromodulatory switching — Marder's principle |
| 17 | Stochastic synapses — noise and resonance |
| 18 | LIF vs AdEx — does topology dominate? |
The FlyWire Connectome
The publicly available FlyWire connectome provides the data substrate: 139,255 neurons, 50 million synapses, 8,453 cell types. BRAVLi loads, parses, and analyses this connectome at every scale — from individual neuropil regions to whole-brain connectivity patterns.
From Anatomy to Dynamics
The lesson sequence traces a deliberate arc:
- Structure (lessons 00–06) — What is the brain made of? Regions, cell types, neurotransmitters, morphologies, geometry.
- Connectivity (lessons 08–09) — How are neurons wired? Synaptic connectivity matrices, physiological parameters, neurotransmitter-receptor matching.
- Dynamics (lessons 10–18) — What does it do? LIF simulation, Brunel phase diagrams, neuromodulatory state switching, stochastic synapses, learning rules. The question: can this wiring diagram come alive?
Heritage
BRAVLi inherits ideas from the Blue Brain Project's cell atlas pipeline, the bravlibpy circuit analysis library, and the circuit-factology measurement framework — adapted for the publicly available FlyWire connectome. It is the neuroscience domain application within the MayaLucIA personal scientific computing environment.
Technical Stack
Python (NumPy, SciPy, navis, neuprint-python) · Org-mode literate programming · FlyWire connectome · Makefile (tangle + test) · LIF / AdEx neuron models · STDP learning rules