Neuromorphology — 3D neurons & astrocytes, and a fly connectome
▸ Reconstruct the shapes in 3D → — a standalone WebGL viewer that streams the electron-microscopy meshes and rebuilds each cell in your browser; rotate an astrocyte, or show the whole neuron cluster at once, coloured by neurotransmitter.
▸ Query it in the playground →
— the same cells as a .rete graph, the 3D meshes rendering inline in the
results table.
This is an experiment: a demonstrator built from two of the neuroscience datasets harvested into rete, and the point it makes is about file formats as much as neurons — the difference between a lossless, queryable graph and a lossy 3D render preview, kept as two layers instead of one.
What's in it
Just 30 cells, chosen to be interesting rather than large:
- 3 astrocytes from the Blue Brain Neuro-Glia-Vascular reconstruction
(Zisis et al. 2021), traced from FIB-SEM electron microscopy. Each one
links (
schema:contentUrl) to a Draco-compressed.glbsurface mesh that the playground renders inline — click 🧊 to rotate it. Astrocytes have a fine "spongiform" morphology, sheet- and leaflet-like processes that sit below the resolution of light microscopy; these EM meshes show detail that the tracing skeletons in a database like NeuroMorpho simply cannot capture. - 27 neurons from the Janelia hemibrain connectome — a connected cluster
of the Drosophila mushroom body, the fly's learning-and-memory circuit.
Each carries its cell type and predicted neurotransmitter, and they are
joined by 30 real weighted synaptic edges. The synapse count on each
connection is attached to the edge itself with RDF-star
(
<< ?a neuro:connectsTo ?b >> neuro:weight ?w), so an edge is a first-class thing you can query, not a flattened triple.
The point: lossless graph + lossy preview
The astrocyte meshes are enormous — one is 10 million triangles, ~400 MB as raw OBJ. That geometry is not what belongs in the graph. So the dataset keeps two layers:
- the
.reteholds every fact losslessly — the cells, their types, neurotransmitters, morphometrics, the connectome, and the URL of each mesh; - the
.glbis a lossy preview for rendering only: decimated to 15 % of its triangles and Draco-quantised (a few-nanometre grid, well below the imaging resolution), ~8–17 MB instead of hundreds.
The full-resolution OBJ meshes and SWC skeletons stay as the analytical source
of truth. Draco-GLB is a wonderful delivery format for a browser, but it is
lossy and it cannot express a graph — so it rides alongside the .rete, never
replaces it.
Try it
Open the playground and run the built-in examples:
- 🧊 The astrocytes in 3D — returns each astrocyte with its mesh URL; click 🧊 to open the EM reconstruction and rotate it.
- Heaviest synapses in the circuit — reads the RDF-star edge weights; the giant modulatory neurons DPM and APL dominate.
- Neurons by neurotransmitter — the acetylcholine / GABA balance of the mushroom body.
- The connectome cluster as a network — switch Output to Graph to see the wiring draw itself.
Licence & credit
Mixed, and published here as a non-commercial research demonstrator with attribution:
- astrocytes — CC BY-NC-SA 4.0, © BBP/EPFL; cite Zisis et al. 2021, Digital Reconstruction of the Neuro-Glia-Vascular Architecture, Cerebral Cortex 31(12):5686–5703, doi:10.1093/cercor/bhab254;
- hemibrain neurons — CC BY 4.0; cite Scheffer et al. 2020, A connectome and analysis of the adult Drosophila central brain, eLife.
Both licences and the citations travel inside the file's dataset card
(rete card).