Illustrative, uncalibrated movement. A separate simulation archive supplies motor spikes. Sensory input is prescribed; contact does not feed back.
Loading the fly geometry…
Run settings
Sugar 100 Hz; Intact; trial 1; 0.25× speed.
Trial 1 of 30
Precomputed simulation archive
A is intact; B is the selected disconnection. Matching trial numbers do not establish paired inputs between these archive runs.
Drag to rotate. Mouth detail shows the same pose.
At the end of this replay
Motor neuron spikes MN9 and MN6 · left and right · one simulated trial
Each tick is one simulated spike. Neural activity: 0–1 s; later movement is assumed relaxation.
Blue/purple: MN9. Brown: MN6. Drive blocks preserve these saved events.
Input and motor activity
Neural input is prescribed. Touching the droplet is a separate geometric event.
Motor markers brighten for 40 ms after saved spikes; sensory spike times are not exported.
Movement model
Test the movement rule
These controls remove motor drive while retaining the saved spike traces. MN9 controls the main reach and head coupling; MN6 controls the independent distal joints. Parameters are fixed engineering assumptions.
Contact timing
Contact is a sphere-to-triangle surface test. Loss uses mean MN6 activity, so it is not a side-specific pumping model.
Methods & sourcesMotor activity, movement assumptions and reuse
This is a replay from a separate simulation archive, separate from the paired Roundup experiment in Connections. Its assumed, uncalibrated body mapping does not measure an animal eating, predict swallowed volume or feed sensory feedback into the brain.
The signal display follows prescribed input into a saved brain simulation, then saved motor output into the assumed body mapping. Motor markers brighten for 40 ms after saved spikes. This replay export contains motor spike times, not sensory spike times. The paired comparison in Connections uses a separate dataset.
The separate simulation archive contains sugar-input (100 Hz) and water-input (200 Hz) runs, three circuit conditions and 30 repeats per condition. The neural model is from Shiu et al. (2024). MN9 and MN6 come from the same saved file and trial. Matching repeat numbers across conditions do not establish paired inputs.
MN9 is mapped to the main reach and head coupling; left and right MN6 signals drive the respective distal source joints. The source calls those parts labrum_left and labrum_right; their correspondence to feeding anatomy remains unqualified. No movement parameters were fitted to animal videos. The neural interval is 0–1 s, followed by assumed actuator relaxation.
Geometric mouth–droplet contact gates an illustrative volume change. This is not fluid physics, pumping, swallowing or a closed sensory loop. The body is not registered to the FlyWire neuron coordinates shown in the anatomy view.
FlyBody geometry is redistributed under Apache 2.0; Three.js and the source neural model retain their MIT licenses. Source and processing details.