Neurotransmitter sign and source-column errors in Drosophila connectome simulation pipelines: an audit of MaleCNS v1.0

Abstract

Spiking simulations of the Drosophila connectome routinely assign each synapse an excitatory or inhibitory sign by mapping the neurotransmitter (NT) predicted for the presynaptic body to a polarity. Using the recently released MaleCNS v1.0 whole-central-nervous-system connectome (166,700 neurons; 25,582,938 intra-dataset edges), we audited this mapping and found four systematic error classes. First, the per-body predicted_nt column mislabels all 4,064 Kenyon cells as dopaminergic, whereas the curated consensus_nt calls them cholinergic; a naive three-class mapping on predicted_nt would reclassify 884,631 of the 885,611 Kenyon-cell output edges as modulatory, inflating the modulatory count from 435,541 to 1,531,095. Second, histamine—the inhibitory photoreceptor transmitter acting through chloride channels—is treated as excitatory, mis-signing 89,723 edges, a 3.7-fold undercount of the histaminergic pool in predicted_nt (2,126 vs 7,891 neurons). Third, dopamine, serotonin and octopamine (541 neurons; 435,541 edges, including 110,450 PAM→Kenyon-cell contacts) fall through to a default excitatory value, so G-protein-coupled neuromodulation is delivered as fast synaptic drive. Fourth, propagated annotation errors affect gustatory receptor neurons and a recurring claim that PEN/PEG compass neurons are GABAergic. We provide a corrected source column, a three-class (excitatory/inhibitory/modulatory) mapping, and a reproducible audit, and show that the reclassification is exact and bit-for-bit reversible.