MaleCNS v1.0 Fly Connectome, Visualized Neuron by Neuron

A connectome is the complete wiring diagram of a nervous system. This console tells that story in chapters, from the whole animal down to a single flash on the retina, on the real released data. The reference connectome is the male Drosophila central nervous system, MaleCNS v1.0, published by the FlyEM team at HHMI Janelia with Cambridge, the MRC Laboratory of Molecular Biology and Google Research, released 8 June 2026 under CC-BY 4.0.

Nothing on this page quotes a number we did not recompute from the released files. Every asset and every count is produced by a documented pipeline that reads only the released files: annotations, the connection table, the synapse table and the neurotransmitter predictions, all downloadable without credentials from male-cns.janelia.org. The steps are listed in The engine chapter below.

// shown in the engine chapter

def scan_weights(path, body_ids, ...):
    # 151,856,684 rows, streamed in 16 M row batches
    for batch in table.to_batches(max_chunksize=16_000_000):
        ...

def run_sim(graph, seeds, ...):
    # leaky integrate and fire on the signed graph
    for step in range(steps):
        inflow = graph.dot(spikes)
        v = leak * v + inflow + drive
        fired = v > threshold
        ...

What this is and what it is not

It is the real wiring of one male fly’s central nervous system, baked from the released data, plus our own simulation pass over that wiring. It is not a recording of a brain, and it is not a biophysical model.

The honest weak points, stated plainly:

• The simulation is a graph traversal with spike timing. Each neuron integrates signed input, leaks, fires past a threshold of 0.1 of its total input weight, and goes refractory. There are no ion channels, no dendrites, no voltage-gated dynamics. The Single Neuron Console runs one cell that has them. The latency numbers in chapter six are properties of that traversal on the released graph, not biological measurements.
• The point cloud is synapse locations, not morphology. One dot per 4 um cube of presynaptic sites. It looks like the animal because the synapses fill the neuropil. The traced anatomy overlay adds 300 sampled, simplified neuron skeletons from the released dataset. These are measured shapes, not every neuron or every synaptic path.
• The matrix chapter draws the strongest connections only. The full graph has 151.9 million edges above one contact. Shipping all of them to a browser would dominate the page weight, so the console draws the 1.1 million at or above twenty contacts and states the cut. The counts in the right panel are the full counts, recomputed.
• One animal, one sex. MaleCNS v1.0 is a male. The female brain (FlyWire) is a different dataset with different annotations.

The other three tabs carry less, and each says so.

• The worm has the complete wiring of the hermaphrodite, chemical synapses and gap junctions, but Cook et al. 2019 released connections, not positions. The neuron positions come from a model of the body, so the anatomy chapter is marked as a model. The touch replay leaves the gap junctions out, because the release gives them no sign and no gain could be set honestly.
• The human slab (H01) is one cubic millimetre of temporal cortex, a small sample of one cortical area. The console draws 16,087 somas classed as neurons, each at its measured position with its recorded layer. It shows how many excitatory and inhibitory synapses each neuron receives, from the release’s per-segment counts, but not which neuron each synapse comes from. The full synapse table is 159 GB, and most incoming axons in the volume belong to cells whose bodies lie outside it. The counts come from automated segmentation, so a dendrite cut off from its cell is missed, and the 97 neurons that share a segment with another cell body, the 268 with no counted segment and the 100 whose segment receives no synapse at all are left grey.
• The mouse volume (MICrONS) is one cubic millimetre of visual cortex. The console draws the 50,948 core neurons of release 1412, each with its type and, for excitatory cells, its layer. The release leaves out about 20,000 neurons at the edge of the volume. The wiring is shown as degrees and as a table of connection probability between cell groups, computed from all 7.5 million connected pairs. Only 1,871 axons are proofread and the rest are cut short at the edge, so those probabilities are lower bounds. The released two-photon recordings are not shown yet.

Get the data. MaleCNS v1.0 and the other releases

Every dataset in this console is public, and none of it needs an account. The console ships a reduced copy of each one. The full releases are at the addresses below.

MaleCNS v1.0 (the fly). Licence CC-BY 4.0. Release page male-cns.janelia.org, downloads on its download page. The connectome tables are Apache Arrow Feather files in gs://flyem-male-cns/v1.0/connectome-data/flat-connectome/:

The microscope images in the slice chapter come from the released image volume gs://flyem-male-cns/em/em-clahe-jpeg, contrast-normalised FIB-SEM in Neuroglancer precomputed format. The console shows twelve cross sections at 1 micrometre per pixel and one close-up at 16 nanometres per pixel, cut from that volume by our build script.

The same page lists the EM image volumes, the v1.0 segmentation, neuron skeletons in SWC and neuroglancer formats, the neuropil regions, and a neo4j database for neuPrint (dataset male-cns:v1.0). To read the weights table in Python, pandas.read_feather is enough. This console used four of the tables. They are the annotations, the body transmitters, the weights and the synapse partners.

C. elegans, Cook et al. 2019 (the worm). Distributed by the OpenWorm C. elegans Connectome Toolbox under the MIT licence, and installable as the Python package cect. The console used the corrected SI 5 tables of July 2020, chemical and electrical connections both.

H01 (human cortex). Licence CC BY 4.0. Release page h01-release. The console used the soma table of release c3, which gives every soma’s position, cell class and layer, and the per-segment count tables tables/segments/counts*.csv.gz, 452 files and 3.7 GB, which give the incoming and outgoing excitatory and inhibitory synapses of every segment. The console keeps the rows of the segments that hold a neuron’s cell body.

MICrONS minnie65 (mouse cortex). Licence CC BY 4.0. Release page microns-explorer.org, source of record on BossDB. The console uses release 1412 as prepared by M. W. Reimann in microns_mm3_connectome_v1412.h5, 761 MB, on Zenodo under CC BY 4.0. It holds the cell types, the proofreading status, the soma positions and one edge per connected pair with its synapse count.

Why it belongs here

The Neutral Core claims substrate independence: the same causal organisation can run on a different physical base. Connectomics is where that claim meets data. A complete nervous system is now small enough to download, and this page exists so the claim can be looked at directly, chapter by chapter, without a credential or a cluster.

The fly is the reference organism for this site’s emulation coverage. The FlyWire whole-brain connectome post covers the female brain. The Eon Systems fruit fly emulation post covers what happens when a connectome drives a simulated body. This console holds the male CNS data itself.

Sources

• MaleCNS v1.0, FlyEM Team, HHMI Janelia, released 8 June 2026, male-cns.janelia.org, CC-BY 4.0. Preprint: bioRxiv 2025.10.09.680999.
• C. elegans connectivity: White et al. 1986; Cook et al. 2019, wormwiring.org.
• MICrONS mm3: microns-explorer.org, The MICrONS Consortium, Nature 2025 collection. Release 1412 core connectome as prepared by Reimann, M. W. (2025), Connectomics of (part of) the MICrONS mm3 dataset, version 3.0.0, Zenodo, doi:10.5281/zenodo.16744240, CC BY 4.0.
• H01: Shapson-Coe et al. 2024, “A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution”, Science 384, eadk4858. CC BY 4.0.
• Whole brain counts in the scale chapters are published estimates, not data this console ships. Human: Azevedo et al. 2009, “Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain”, J. Comp. Neurol. Mouse: Herculano-Houzel, Mota and Lent 2006, “Cellular scaling rules for rodent brains”, PNAS. Honeybee: Menzel and Giurfa 2001, “Cognitive architecture of a mini-brain”, Trends in Cognitive Sciences.