Google Research and HHMI Janelia Complete First Adult Male Fruit Fly CNS Connectome

**Google Research and collaborators at HHMI Janelia Research Campus have completed the first finalized connectome of an entire adult male *Drosophila melanogaster* central nervous system. The release covers the central brain, both optic lobes and the ventral nerve cord, containing 166,691 neurons and 125 million synaptic connections—the largest brain map by neuron count to date.** Google Research announced the release on September 3, 2026, with the study published in Cell.\n\nThe map extends beyond a brain-only atlas. The ventral nerve cord is analogous to a spinal cord, while the optic lobes perform functions comparable to a mammalian retina. Connecting these regions allows researchers to trace sensory inputs, including visual, auditory and olfactory signals, toward motor outputs across the central nervous system. The Janelia project describes the dataset as fully proofread and annotated, with synaptic-resolution comparisons between male and female flies.\n\nThe achievement follows a series of smaller connectomics milestones. In 2020, the hemibrain map covered 25,000 neurons in roughly half of a female fly brain. In 2024, a separate whole adult female brain map reconstructed 139,255 neurons and 54.5 million synapses. The new male CNS resource complements those datasets by adding a complete male brain-and-nerve-cord view, making sex-linked circuit differences easier to examine.\n\nAt imaging scale, researchers used seven enhanced focused ion beam scanning electron microscopy systems over 13 months to produce a 160-teravoxel volume at 8×8×8 nanometer isotropic resolution. AI-based flood-filling networks automatically segmented neurons, while automated detection identified 46 million presynapses and 312 million postsynaptic densities with average precision and recall of 0.82 and 0.81, respectively. The initial segmentation then required an estimated 44 person-years of human proofreading.\n\nThe biological analysis identified 11,691 cell types. Comparing male and female brain connectomes, the study found 7,205 isomorphic types, 114 dimorphic types, 262 male-specific types and 69 female-specific types. Sex-specific and dimorphic neurons were concentrated in higher-order brain centers, while sensory and motor regions were largely shared. The analysis also identified circuit switches that reroute sensory information into different behavioral pathways, helping explain how similar inputs can produce different responses.\n\nResearchers have already applied the resource to visual, taste and social-behavior circuits. One analysis found that visual processing extends deep into the brain and involves more than half of its neuron types; another traced taste signals from sensory neurons across the body to circuits controlling feeding, locomotion, hormone release and courtship. The earlier hemibrain dataset enabled hundreds of research findings and helped other teams begin working with connectomes, according to coverage of the project’s development.\n\nThe release is also designed for practical reuse. Researchers can browse cell types, compare male and female circuits, query connectivity through NeuPrint, inspect volumes in Neuroglancer, and access the data through Python and R tools. The download package includes bulk connectivity tables, neuron skeletons, annotations and neurotransmitter predictions under a CC-BY license. The trade-off is scale: the listed synapse-point table is 12.7 GB and the synaptic-partner table is 6.8 GB, while the recommended Python workflow requires a NeuPrint account and API token.\n\nThe dataset is primarily a structural resource, listing electron-microscopy volumes, segmentations, synapses, skeletons, annotations and neurotransmitter predictions. It does not provide the live neural-activity recordings described separately for the team’s zebrafish work, and the need for extensive human proofreading shows that AI has reduced but not eliminated manual reconstruction. Google Research says the same methods are now supporting work on larval zebrafish, adult *Danionella* and portions of the mouse brain, while mapping the human brain’s roughly 86 billion neurons remains beyond current capabilities.