Does the Wiring Matter? Testing a Connectome-Extracted Escape Circuit Against Its Own Randomized Control
Authors/Creators
-
1.
Sukkur IBA University
- 2. Independent Researcher
Description
Connectomics datasets now provide synapse-resolution wiring diagrams for entire insect nervous systems, raising
a question that is easy to pose but rarely tested directly: does the specific wiring these datasets reveal confer a
functional advantage over a structurally-matched random control, when deployed as an actual real-time controller
— or does circuit scale and connectivity pattern alone explain performance, regardless of whether the wiring is
biological? We address this with a case study: a looming-avoidance circuit (four visual projection neuron types,
their strongest interneuron relays, and the descending neurons they drive) is extracted from the male Drosophila
melanogaster CNS connectome (male-cns:v1.0) and deployed as a real-time controller in two interactive evasion
tasks — a planar (2D) task and a full 3D free-flight task — where it must steer a simulated agent away from falling
obstacles using only its real, synapse-weighted output. We compare the extracted circuit against a degree-and-weightmatched random rewiring of itself, isolating the effect of specific connectivity from generic circuit-scale effects. In
the 2D task, the real circuit shows a large, statistically robust advantage over its randomized control (1.1% vs. 19.2%
obstacle-contact rate over 448 trials per condition; non-overlapping 95% confidence intervals; consistent across every
tested control-gain setting). In the 3D task, this advantage is not detectable (Welch’s t-test, p > 0.05 at every tested
setting, n = 24 seeds per condition), which we attribute to the 3D task’s second control axis having no connectomic
basis at all in our extraction — a boundary condition of the extraction, not evidence against the first result. We
discuss this axis-specific pattern, the method’s limitations, and the motivating — but here untested — prospect of
connectome-derived controllers for compute-constrained edge robotics as a direction for future work.
Files
paper (1).pdf
Additional details
Related works
- Is supplemented by
- Software: https://github.com/cook-cxb/flybrain-dodge (URL)
- Software: 10.5281/zenodo.23162131 (DOI)
Software
- Repository URL
- https://github.com/cook-cxb/flybrain-dodge
- Programming language
- HTML , JavaScript , Python , Shell
- Development Status
- Active
References
- Ache, J. M., Polsky, J., Alghailani, S., Parekh, R., Breads, P., Peek, M. Y., Bock, D. D., von Reyn, C. R., and Card, G. M. (2019). Neural basis for looming size and velocity encoding in the Drosophila giant fiber escape pathway. Current Biology, 29(6), 1073–1081. https://doi.org/10.1016/j.cub.2019.01.079
- Jang, H., Goodman, D. P., Ausborn, J., and von Reyn, C. R. (2023). Azimuthal invariance to looming stimuli in the Drosophila giant fiber escape circuit. Journal of Experimental Biology, 226(8), jeb244790. https://doi.org/ 10.1242/jeb.244790
- Klapoetke, N. C., Nern, A., Peek, M. Y., Rogers, E. M., Breads, P., Rubin, G. M., Reiser, M. B., and Card, G. M. (2017). Ultra-selective looming detection from radial motion opponency. Nature, 551, 237–241.
- FlyEM (HHMI Janelia), University of Cambridge Department of Zoology, MRC Laboratory of Molecular Biology, and Google Research. MaleCNS connectome, release male-cns:v1.0. https://male-cns.janelia.org/. Licensed CC-BY.
- Qin, Z., Peng, J., Yue, S., and Fu, Q. (2025). A Bio-Inspired Research Paradigm of Collision Perception Neurons Enabling Neuro-Robotic Integration: The LGMD Case. arXiv preprint arXiv:2501.02982