Male Fruit Fly Connectome Maps 166,691 Neurons Across Brain and Nerve Cord

Google Research, the Howard Hughes Medical Institute’s Janelia Research Campus, and an international group of collaborators have published a synapse-resolution connectome of an adult male fruit fly’s central nervous system. The reconstruction contains 166,691 proofread neurons spanning the central brain, both optic lobes, and the ventral nerve cord—the insect structure commonly compared with a vertebrate spinal cord.

Google describes the resource as the largest brain map to date when measured by the number of proofread neurons. It reports approximately 125 million synaptic connections and 11,691 annotated neuronal types. The accompanying study appeared in *Cell* on September 3, 2026, while Google AI highlighted the work publicly on September 9.

The result is not a simulation of a fly’s mind. It is a structural map showing neurons, their shapes, and the physical connections through which signals may travel. Its immediate value is that researchers can now inspect continuous pathways from sensory inputs, through higher brain regions, and into circuits associated with movement and other bodily actions.

1. What the New Connectome Contains

The MaleCNS dataset reconstructs the central nervous system of a single adult male *Drosophila melanogaster*. Its scope includes the central brain, the visual-processing optic lobes, the neck connective, and the ventral nerve cord.

That continuous coverage matters. Earlier connectomes often mapped the brain or nerve cord separately, requiring researchers to match neurons across different specimens and datasets when studying complete sensory-to-motor pathways. MaleCNS preserves the connection between the brain and nerve cord in one specimen, allowing a circuit to be followed without crossing an artificial boundary between two reconstructions.

The project follows several earlier milestones. A 2020 “hemibrain” reconstruction covered approximately 25,000 neurons and 21 million connections in part of a female fly brain. In 2024, the FlyWire Consortium published a complete adult female brain containing 139,255 neurons and 54.5 million synapses. A separate 2026 Brain and Nerve Cord dataset subsequently united a female brain and ventral nerve cord, with 155,916 proofread or roughly proofread neurons reported in that resource.

MaleCNS therefore is not the first complete fruit fly brain or the first brain-and-cord reconstruction. Its distinct contributions are a finished male central nervous system, the highest published count of proofread neurons in a brain map, and a reference that can be compared with existing female connectomes at individual-synapse resolution.

The word “complete” also requires qualification. The authors use it to describe the reconstruction’s anatomical coverage and their accounting of essentially all traceable neurons, rather than to claim that every biological detail is error-free. A small number of cells could not be reconstructed because of sample-edge artifacts or segmentation problems.

The study reports that 94% of presynaptic sites and 42% of postsynaptic sites in neuropil regions were associated with proofread neurons. For 40.1% of detected synaptic connections, both sides belonged to proofread neurons. These rates are comparable to or better than those of earlier large connectomes, but they make clear that “complete wiring diagram” does not mean a perfect inventory of every synapse.

2. How AI and Human Proofreading Produced the Map

The reconstruction began with enhanced focused ion beam scanning electron microscopy. Seven microscopy systems operated for 13 months to image a volume of 0.082 cubic millimeters at an isotropic resolution of eight nanometers. The resulting dataset contained approximately 160 teravoxels.

Computational systems aligned the microscopic images into a coherent three-dimensional volume. Google’s flood-filling networks then performed initial neuron segmentation. These convolutional neural networks start from a location in an image and predict which surrounding pixels belong to the same biological object, enabling long, intertwined neural processes to be separated computationally.

Algorithms also detected likely synapses, nuclei, and neurotransmitter properties. These predictions converted the electron-microscope volume into candidate neuron shapes and connections that specialists could inspect.

AI did not eliminate manual reconstruction. Automated segmentation can incorrectly merge two neurons or break one neuron into multiple fragments, and those errors can distort every downstream connectivity analysis. Human experts therefore reviewed the generated structures, corrected segmentation errors, and assigned anatomical and cell-type annotations.

The paper estimates that proofreading the initial segmentation required about 44 person-years of work. Researchers checked every automatically segmented fragment with more than 100 synaptic connections and used detected cell nuclei as an additional accounting mechanism. Of 141,780 detected neuron-associated nuclei, 98.9% were directly attached to a proofread neuron; the team estimated that almost all of the remainder had also been covered through its fragment-based review.

This division of labor is the principal AI result behind the announcement. Machine learning made it feasible to turn a nanometer-scale image volume into candidate three-dimensional neurons, but biological experts remained responsible for verifying the structures and interpreting what they represented. The dataset is consequently an AI-assisted scientific reconstruction, not the output of an autonomous discovery system or a large language model.

3. Male and Female Brains Can Now Be Compared Synapse by Synapse

The male map enables the first comprehensive synaptic-resolution comparison between male and female fruit fly brain connectomes. Researchers matched 7,205 neuronal types that were structurally similar between the sexes, classified 114 shared types as sexually dimorphic, and identified 262 male-specific and 69 female-specific types.

“Sex-specific” types appear in one sex but lack a direct counterpart in the other reference connectome. “Dimorphic” types exist in both but differ in morphology or connectivity. Together, these categories account for 4.8% of neurons in the male central brain and 2.4% in the female central brain.

The relatively small number of visibly different neurons understates their network reach. The study estimates that about 18% of neurons in the male brain and 8% in the female brain participate in connections that differ significantly between the two datasets. Sex-specific and dimorphic neurons also connect disproportionately with one another, creating concentrated regions where modest structural changes can influence wider circuits.

These differences were not distributed uniformly. The researchers found that sensory and motor neurons at the periphery were largely similar between males and females, while sex-specific and dimorphic neurons were concentrated in higher-order brain centers. Some circuits act as switches, routing shared sensory information toward different downstream pathways.

The annotations include expression information for the transcription factors *fruitless* and *doublesex*, both associated with sexual differentiation in flies. This gives researchers a way to relate genetic regulation to cell identity, circuit architecture, and experimentally observed behaviors such as courtship and aggression.

The connectome alone cannot prove that a particular connection causes a behavior. It instead narrows the experimental search space: scientists can identify candidate neurons or pathways, manipulate them in living flies, and test whether the predicted behavioral effect occurs.

4. An Open Research Resource, Not Just a Published Image

MaleCNS version 1.0 was released on June 8, 2026, before the peer-reviewed paper appeared in September. The dataset is licensed under CC BY and can be explored through Neuroglancer, neuPrint, Clio, and a dedicated cell-type explorer.

Researchers can also download the segmentation volumes, neuron skeletons, annotations, neurotransmitter predictions, synaptic locations, partner tables, and the complete connection graph. Janelia provides programmatic access as well as a downloadable Neo4j database backing the public neuPrint instance.

These access options make the connectome useful beyond the group that constructed it. A laboratory studying vision can retrieve neurons downstream of particular photoreceptors; a behavioral researcher can inspect paths between sensory neurons and motor outputs; and computational neuroscientists can analyze the entire network without repeating the imaging and reconstruction process.

Three companion studies released with the main paper apply the resource to visual processing, taste, and social behavior. One mapped a path from R1–R6 visual neurons toward the DNg13 motor neuron, including a male-specific neuron associated with the “love spot,” a specialized visual region involved in courtship. Other analyses trace taste information from sensory neurons across the body into circuits related to feeding, movement, hormone release, and courtship.

The map also supplies standardized cell types that can be matched against light-microscopy images, genetic driver lines, physiological recordings, and previous connectomes. That interoperability is important because a structural prediction becomes experimentally useful only when researchers can identify and manipulate the corresponding cells in living animals.

5. What the Map Does—and Does Not—Explain

A connectome records physical wiring, not the complete state of a functioning nervous system. It does not directly capture moment-to-moment electrical activity, neuromodulation, learning, synaptic plasticity, or the effects of the fly’s history and internal condition. The map therefore cannot by itself reproduce a fly’s behavior or establish what an individual neuron “means.”

The reconstruction also represents one male specimen. Comparisons with female datasets include genuine sex-linked differences, ordinary variation between individual animals, and possible differences introduced by imaging or reconstruction methods. The researchers used morphology, connectivity, and known genetic markers to make conservative matches, but additional male and female connectomes will be needed to measure population-level variation.

Its practical contribution is more bounded and testable. MaleCNS gives neuroscientists a common, inspectable reference for forming hypotheses about how sensory information reaches motor and behavioral circuits. It also demonstrates that AI-assisted segmentation, large-scale electron microscopy, and structured human proofreading can produce a whole-central-nervous-system map containing more than 166,000 verified neurons.

Frequently Asked Questions

Is this the first complete fruit fly brain map?

No. Researchers published a 139,255-neuron connectome of an adult female brain in 2024. MaleCNS is the first finished connectome of an entire male fruit fly central nervous system and the largest brain map by proofread neuron count.

Does the dataset include the fly’s spinal cord?

It includes the ventral nerve cord, an insect structure analogous to the vertebrate spinal cord, together with the intact neck connection to the brain.

Did AI map the nervous system without human assistance?

No. AI systems aligned images, segmented neurons, and detected candidate synapses, but specialists spent an estimated 44 person-years proofreading structures and adding annotations.

Is the connectome a working simulation of a fruit fly?

No. It is a structural wiring diagram. Simulating behavior additionally requires models of neural activity, synaptic dynamics, sensory input, muscles, and the body’s interaction with its environment.

Can researchers download the data?

Yes. Janelia provides downloadable volumes, annotations, neuron skeletons, connectivity tables, synapse data, and database exports under a CC BY license.

Sources

• Original Google AI announcement on X
• Google Research: A connectomics milestone—Mapping the complete male fruit fly brain
• Google Research publication record: Sexual dimorphism in the complete connectome of the Drosophila male central nervous system
• Full research article in PubMed Central
• HHMI Janelia: Male CNS Connectome
• MaleCNS version 1.0 download documentation
• Nature: Neuronal wiring diagram of an adult brain
• Nature: Distributed control circuits across a brain-and-cord connectome
• Live Science: Male fly central nervous system map enables direct comparison with the female brain