FlyWire skeleton cleanup maps where neurons receive and send signals

A team analyzing the FlyWire connectome of the adult fruit fly brain has released a standardized, quality-controlled skeleton resource covering 139,273 neuronal reconstructions, and used it to uncover a consistent pattern: in nearly every region of the fly brain, the branches where neurons receive signals sit farther out on the dendritic tree than the branches where they send them. The work, published in Brain Informatics, addresses a persistent bottleneck in modern neuroscience—turning massive automated reconstructions into reliable anatomical data.

Skeletons are simplified line drawings of neurons that preserve branching structure while discarding volume, making them essential for comparing cell shapes and wiring diagrams at brain-wide scale. But automated skeletonization of dense electron microscopy data can introduce local structural and topological artifacts, particularly around the soma, the neuron’s cell body. Such errors can distort measurements of branch order, the number of bifurcations separating a point from the root of the tree.

To tackle this, Longxiao Yuan, Zhiyan He and Hanchuan Peng systematically assessed all 139,273 skeletons in FlyWire version 783 and built a human-in-the-loop refinement framework. The procedure corrected soma-associated artifacts, standardized branching topology, and integrated the positions of synapses onto the cleaned skeletons. The refinement was conservative: the median normalized NBLAST similarity, a standard score of how closely two neuronal shapes match, was 0.992 between original and refined skeletons, and every neuron retained its correct Top-1 identity match. In other words, the cleanup removed artifacts without reshaping the cells or scrambling their classifications.

With the standardized resource in hand, the researchers examined where synapses fall along the branching hierarchy of neurons across the fly brain’s neuropils, the densely interwoven regions where synaptic connections occur. Nodes associated exclusively with postsynaptic sites—the receiving ends of chemical synapses—sat at higher mean branch orders than nodes associated exclusively with presynaptic sites, the transmitting ends, in 39 of the 43 neuropils examined.

The team also confirmed that this asymmetry was not an artifact of how branch points were counted. When multifurcations, points where more than two branches meet, were left unbinned rather than converted into binary splits, the pattern held. The consistency across counting schemes suggests the effect reflects genuine biological organization rather than a quirk of the analysis pipeline.

Because the skeletons, synapse annotations and refinement procedures are standardized, the resource can serve as a common substrate for brain-wide analyses of neuronal morphology and synaptic topology, allowing laboratories to compare measurements across the full connectome rather than within small, idiosyncratically processed subsets. The authors position it as infrastructure for the next wave of connectomic studies, in which questions about wiring must be asked at the scale of the entire brain.

Subject of Research: Standardized, quality-controlled FlyWire neuronal skeletons and their neuropil-specific synaptic topology

Article Title: Standardized FlyWire skeletons reveal neuropil-specific organization of synaptic topology

Article References: Yuan, L., He, Z., & Peng, H. (2026). Standardized FlyWire skeletons reveal neuropil-specific organization of synaptic topology. Brain Informatics. https://doi.org/10.1186/s40708-026-00344-8

Image Credits: AI Generated

DOI: 10.1186/s40708-026-00344-8

Keywords: FlyWire, connectomics, skeletonization, Drosophila, synaptic topology, branch order, neuropils, NBLAST, morphology, quality control, Standardized, skeletons