What Is MaleCNS? The First Whole-CNS Wiring Diagram of a Male Fruit Fly, and How It Differs from FlyWire
What Is MaleCNS? The First Whole-CNS Wiring Diagram of a Male Fruit Fly, and How It Differs from FlyWire
If you've looked at the recent wave of "fruit fly brain plays a game" projects, you'll notice that nearly everything built after September names the same dataset: MaleCNS.
It stands for Male Central Nervous System: the wiring diagram of a male fruit fly's entire central nervous system. The short version:
- It's the first fly wiring diagram with the brain and the ventral nerve cord together. The ventral nerve cord is the fly's spinal cord, home of the leg and wing motor neurons. FlyWire covers only the brain.
- Scale: 166,700 neurons and 11,710 neuron types, all proofread and annotated.
- Timeline: v0.9 in October 2025, v1.0 on June 8, 2026, and the paper in Cell in September 2026.
- Licence: CC BY 4.0, commercial use allowed with attribution. FlyWire is CC BY-NC 4.0, no commercial use.
- It's an entirely separate effort from FlyWire: a fly of the other sex, a different microscope, a different AI segmentation.
- Male and female fly brains are mostly the same. The paper finds sex differences in only a small set of cell types, concentrated in higher brain centres.
MaleCNS vs. FlyWire
MaleCNS is a collaboration between Janelia's FlyEM team, the University of Cambridge's Department of Zoology, the MRC Laboratory of Molecular Biology and Google Research. The numbers below come from its paper (Cell 2026 and the bioRxiv preprint) and the two original FlyWire papers:
| FlyWire (FAFB) | MaleCNS | |
|---|---|---|
| Fly | One female | One male |
| Coverage | Brain (incl. optic lobes) | Brain + optic lobes + ventral nerve cord |
| Microscope | Serial-section transmission EM | Enhanced focused ion beam SEM (eFIB-SEM) |
| Cutting | 7,062 sections, 35–40 nm each | "Hot-knife" cut into 20 µm slabs, then milled and imaged layer by layer |
| Voxel | 4 × 4 × 40 nm | 8 × 8 × 8 nm, equally fine in all three directions |
| Imaging | 2 custom microscopes, ~16 months | 7 microscopes in parallel, 13 months, 160 teravoxels |
| Segmentation | Princeton's boundary-detecting convolutional nets | Google's flood-filling networks (FFN) |
| Neurons | 139,255 | 166,700 |
| Cell types | 8,453 | 11,710 |
| Proofreading | ~33 person-years: research community, professionals, citizen scientists | ~44 person-years: 29 expert proofreaders over 3 years |
| Synapses | ~130 million (54.5 million between proofread neurons) | 46 million presynapses connected to 312 million postsynaptic sites |
| Licence | CC BY-NC 4.0 (non-commercial) | CC BY 4.0 (commercial use allowed) |
Two rows deserve unpacking:
1. The imaging strategies are opposites. FlyWire cut the brain into thousands of ultra-thin sections and imaged each; MaleCNS cut only a few dozen 20-micron "slabs", then in a focused-ion-beam microscope repeatedly milled away a thin layer of the surface and imaged what was revealed. That gives 8 nm resolution in every direction and cleaner volumes, but each machine is slow — hence seven running at once. The ventral nerve cord alone was cut into 31 slabs.
2. Synapses are counted differently. Fly synapses are often one-to-many: a single presynaptic site contacts several downstream neurons' postsynaptic sites. MaleCNS reports 46 million presynapses connected to 312 million postsynaptic sites, while FlyWire's "54.5 million synapses" counts each presynapse–postsynapse pair. So the two numbers can't be compared directly. The "25 million connections" you'll see in project READMEs is yet another unit — connected neuron pairs. I listed these traps in a table in the community projects roundup.
Why the ventral nerve cord matters
However complete FlyWire is, it stops at the neck. A decision to turn left or take off travels through about 1,300 descending neurons to the ventral nerve cord, whose circuits and motor neurons turn it into leg and wing movements. Without the ventral nerve cord, that stretch has to be filled in by hand.
That was the weak point of Eon Systems' March 2026 "uploaded fly" demo. Eon acknowledged it: "We can't trace the actual motor neurons because the body was not scanned." So it read a few descending neurons and handed them to controllers trained into the body model (see Was a Fruit Fly Brain Really Uploaded?).
MaleCNS fills that gap. The paper traces complete pathways from sensory neurons to motor neurons, and finds that the neck connective is a key bottleneck: descending neurons limit information flowing from brain to nerve cord, ascending neurons limit the reverse — and these neurons integrate information rather than merely relaying it.
The community has started using it. The macOS desktop pet DesktopFly extracts a 1,045-neuron leg locomotor circuit from MaleCNS, including 220 real leg motor neurons, and lets simulated motor-neuron spikes drive the joints directly — impossible with FlyWire alone.
How male and female brains differ
This is the theme of the MaleCNS paper, and the first comparison of two complete male and female brains at synaptic resolution.
The paper's findings (Cell 2026): among comparable cell types, 8,069 are isomorphic (the same in both sexes), 138 dimorphic (present in both but shaped or wired differently), 289 male-specific and 71 female-specific. Differences concentrate in higher brain centres, while the sensory and motor periphery is largely the same. Within higher centres, male-specific connections cluster into "hotspots" defined by male-specific neurons or arbours, and dimorphic neurons reroute information between the sexes.
A classic example is the P1 neurons: a male-specific population expressing doublesex (some also fruitless), the transcription factors that set sex, which promote courtship song and aggression.
I recomputed it by brain region. The official FlyWire annotation table (flyconnectome/flywire_annotations, 2026-09-29) now labels every neuron in the female brain as isomorphic, sexually dimorphic or female-specific, based on the MaleCNS matching. Grouped by region:
| Region | Neurons | Not isomorphic | Share |
|---|---|---|---|
| Descending neurons (brain → body) | 1,303 | 109 | 8.37% |
| Central brain | 32,383 | 731 | 2.26% |
| Ascending neurons (body → brain) | 1,750 | 37 | 2.11% |
| Motor neurons | 110 | 2 | 1.82% |
| Optic lobes | 77,541 | 287 | 0.37% |
| Sensory neurons | 16,907 | 6 | 0.04% |
| Whole brain | 139,248 | 1,176 | 0.84% |
("Not isomorphic" includes the table's "sexually dimorphic" and "female-specific" labels plus their "potentially" variants.)
Two findings:
- Over 99% of neurons are the same in both sexes. Sensory neurons are almost identical (6 out of 16,907), matching the paper's "largely isomorphic periphery".
- The highest share isn't the central brain — it's descending neurons. One in twelve descending neurons differs by sex. Intuitively that makes sense: given the same sensory input, a male courts and a female lays eggs, and the difference has to land on the commands sent to the body. The paper's abstract doesn't single this out; I computed it from the annotation table. It's a descriptive statistic, not a causal claim.
The table also marks 3,308 neurons as expressing fruitless or doublesex, 2.4% of the female brain.
How to get and use it
- Website: male-cns.janelia.org — download images, annotations, synapses, neuron skeletons and more, plus online tools for browsing cell types and the catalogue of male–female differences.
- Queries: neuPrint (neuprint.janelia.org, dataset male-cns:v1.0), in the browser or programmatically.
- Licence: CC BY 4.0 — attribution is all that's required, including for commercial use.
- Versions: v0.9 (2025-10-05) was the initial release; v1.0 (2026-06-08) brought only minor proofreading changes and refined annotations, so results based on v0.9 largely still hold.
For whole-brain simulation: the official Shiu 2024 whole-brain LIF model is built on FlyWire and runs on an ordinary computer (see How to Download the Fruit Fly Brain and Run It); community projects such as DOOMFLY, fly-flappy and CyberFly have ported the same LIF approach to MaleCNS — see the community projects roundup.
FAQ
How do MaleCNS, MANC, the hemibrain and FlyWire relate?
All are fly connectomes with different coverage. The hemibrain (2020, Janelia and Google) covers part of the central brain; MANC is the male adult nerve cord connectome; FlyWire (2024) is one female's whole brain; MaleCNS (2026) is one male's brain plus nerve cord, the first dataset with the entire central nervous system.
Why do neuron counts differ between sources?
The preprint says 166,691, the published Cell version 166,700, and some projects 165,122 (counting only neurons marked as traced). The differences come from versions and counting rules; all are around 166,000.
Is FlyWire still useful now that MaleCNS exists?
Yes. Two flies are needed to compare the sexes and to test which connections are stable across individuals. And many existing tools and models, including the Shiu 2024 whole-brain model, are built on FlyWire.
References
- Berg, S. et al. 2026, Sexual dimorphism in the complete Drosophila male central nervous system connectome, Cell. doi:10.1016/j.cell.2026.08.015 (preprint: bioRxiv doi:10.1101/2025.10.09.680999)
- MaleCNS website and release notes: male-cns.janelia.org
- Dorkenwald et al. 2024, Neuronal wiring diagram of an adult brain, Nature. doi:10.1038/s41586-024-07558-y
- Zheng et al. 2018, A Complete Electron Microscopy Volume of the Brain of Adult Drosophila melanogaster, Cell. doi:10.1016/j.cell.2018.06.019
- FlyWire annotations: github.com/flyconnectome/flywire_annotations (commit a83b277, 2026-09-29)
