Bravli

The atlas is where abstract connectivity meets tangible space — 78 neuropil meshes and thousands of neuron skeletons arranged in the coordinate system of a real fly brain. Lessons Covered Lesson 01 — Parcellation The 78 neuropil regions of the Drosophila brain, loaded from FlyWire annotations. Hierarchical tree structure (brain → super-regions → neuropils) with query interface for navigating the anatomy. Lesson 06 — The Volumetric Atlas From point clouds to morphologies: 3D fly brain you can hold in your hands. This lesson loads neuropil surface meshes and neuron skeletons from Zenodo, rendering them as interactive HTML with navis and plotly. ...

Anatomy tells you who connects to whom. Physiology tells you what those connections do. These two lessons build the biophysical parameter database that turns a wiring diagram into a simulation. Lesson 09 — Synaptic Physiology A SynapseModel database covering 6 neurotransmitter types: Reversal potentials — what voltage each synapse drives toward Receptor kinetics — rise time, decay time, conductance amplitude Confidence levels — distinguishing measured values from literature estimates from educated guesses Each parameter carries provenance: where it came from, how reliable it is, what the fly-specific evidence says versus the generic insect value. ...

The FlyWire connectome provides the complete synaptic wiring diagram: every connection between every neuron, annotated with neurotransmitter type and synapse count. This lesson turns that raw edge list into analysable connectivity matrices and pathway maps. Lesson 08 — Connectivity Starting from ~50 million synaptic connections: Edge list loading — parsing the FlyWire synapse table Neurotransmitter assignment — each synapse inherits its presynaptic neuron’s NT identity (ACh, GABA, Glu, 5-HT, DA, OA) Neuropil connectivity matrices — aggregating synapses by source/target neuropil to reveal the coarse wiring diagram Pathway analysis — tracing multi-hop paths between regions Motif analysis — identifying recurring circuit motifs (reciprocal, convergent, divergent) The connectivity matrix is the skeleton on which all simulation and analysis hangs. Get this right, and the dynamics emerge from the wiring. ...

Before building anything, we need a principled way to manage scientific data — datasets that are large, heterogeneous, version-sensitive, and expensive to recompute. The data foundations layer establishes the abstractions that every subsequent lesson builds on. Lessons Covered Lesson 00 — Foundations Datasets, lazy evaluation, and the shape of scientific data management. Introduces the @evaluate_datasets decorator pattern: scientific functions declare what data they need, and the framework resolves, caches, and validates dependencies automatically. ...

The mushroom body (MB) is the fly’s learning and memory center — a microcircuit of ~6,300 neurons that transforms dense olfactory input into sparse, associative representations. It is the first circuit explored end-to-end in BRAVLi, and the subject of a research manuscript. Lessons Covered Lesson 05 — Explore the Mushroom Body Integration: complete factsheet and visualisation of the MB. Kenyon cells (KC), projection neurons (PN), dopaminergic neurons (DAN), and mushroom body output neurons (MBON) — populations, connectivity, and spatial arrangement. ...

The connectome is the hardware. Neuromodulation is the software update that changes what the hardware does without rewiring it. Stochastic synaptic transmission is the noise floor that, paradoxically, can enhance signal detection. Lesson 16 — Neuromodulatory Switching How does the same connectome produce opposite behaviours? Marder’s principle: neuromodulators (dopamine, octopamine, serotonin) alter synaptic gain in a compartment-specific manner, effectively reconfiguring the circuit’s functional connectivity without changing the anatomy. The model implements gain modulation on mushroom body output pathways, switching between approach and avoidance behaviours by changing the balance of appetitive vs. aversive MBON compartments. ...

139,000 neurons on a laptop, from first principles. The simulation engine assembles anatomy (parcellation), connectivity (edge lists), and physiology (cell and synapse models) into a running whole-brain LIF simulation — no Brian2, no NEST, just pure NumPy. Lesson 11 — Whole-Brain LIF Simulation The simulation uses the Shiu et al. formulation: current-based LIF with exponential synaptic conductances, sparse connectivity matrix, and vectorised Euler integration. Key design choices: Pure NumPy — no external simulator dependency; every equation is visible and modifiable Sparse matrices — scipy.sparse CSR format for the 50M-synapse connectivity Stimulus protocols — current injection, optogenetic activation, sensory input patterns Analysis tools — raster plots, firing rate histograms, population synchrony measures Lesson 15 — Brunel Phase Diagram Where does the mushroom body sit in dynamical regime space? The Brunel (2000) framework classifies networks by two axes: synchrony (synchronous vs. asynchronous) and balance (excitation-dominated vs. inhibition-stabilised). By varying external drive and inhibitory gain, we map the MB’s operating point. ...

Where atlas, connectome, physiology, and simulation converge into an interactive exploration tool. The visualisation portal is the feedback loop that closes the scientific cycle: build a model, render it, inspect it, find the gaps, refine. Lesson 04 — Visualization 3D rendering and interactive exploration with navis and plotly: Neuron point clouds — scatter plots of cell body positions, coloured by type Connection matrices — heatmaps of neuropil-to-neuropil connectivity 3D neuropil meshes — surface renderings of the 78 brain regions Interactive HTML exports — explorable in any browser Lesson 12 — The Digital Twin Portal The portal philosophy: a model’s limitations are as informative as its successes. The portal exposes: ...