Whole-brain connectomics of Drosophila reveals a robust, distributed architecture for the suppression of feeding during escape
Survival demands instant prioritization of escape over maintenance. To decode the dynamic logic embedded in the static Drosophila connectome (FlyWire v783), we simulated the conflict between predator-evasion, feeding, and grooming to uncover the logic of this switch. We show that escape is a holistic state defined by distributed robustness: it is reliably triggered by only a fraction of visual inputs (LC4/LPLC2) and induces a system-wide pause on behaviors like feeding and grooming. Within this global suppression, we identify a specialized, fail-safe architecture for arresting feeding. A redundant ensemble of neurons (DNge031/CB0565) targets the premotor center Roundup, effectively disfacilitating motor output. Crucially, this inhibition is modular, anatomically distinct from grooming suppression, and relies on additive logic to ensure suppression even if individual components fail. Our results demonstrate how connectome topology implements robust, survival-critical control through distributed neural architecture.