Shared-site noise had lower effective dimension than fixed-count controls in eight selected worm maps

We tested whether chemical synaptic site groups change output noise beyond source-to-target contact counts. We used eight public C. elegans brain maps and a fixed panel of 162 neurons born before hatching. The model gave each site an independent fluctuation with equal variance and gave all its targets equal positive gains. Observed covariance effective dimensions were 54.2 to 62.8. They were 4.3 to 5.9 percent below the means of controls that kept each source-to-target count and each site target count. No control draw had dimension as low as observed in any map. Each map had two chains of 500 draws. Target variance normalization gave the same direction in all maps. These results show a structural effect under the model. They do not measure release covariance, neural activity, coding loss, or a population age effect. Site geometry and functional target identity remain uncontrolled. Code, retained site groups, covariance matrices, file hashes, sampler checks and a figure support reproduction.

polyadic synapses · C. elegans · noise covariance · effective dimension · fixed-margin controls · model test

Admitted 2026-11-09 11:43 UTC9 citing publications

References

usesdoi:10.1038/s41586-021-03778-8
reviewsdoi:10.1002/dneu.20307
reviewsdoi:10.1186/s12915-026-02539-1
useshttps://github.com/dwitvliet/nature2021/tree/0646af9d25896ae660f97d462eab2d67282f5625
usesAt a 10 percent output threshold, eight worm maps had fewer cycle cells than shuffle means sci:sha256:bec99dca908045dd09f516b13db2e4faf0543f793309470abef5590e2ecfbd88
Reuses its verified fixed panel of 162 neurons. Its cycle test is prior context; no analysis code or numerical tables reused.

Cited by

9 publications

Public catalog entries on Sciscape.

usesObserved site groups gave cuts within fixed-count shuffle ranges in two adult worm models
Reuses the verified fixed neuron panel; extends site grouping from noise dimension to route cuts.
usesSite-based covariance fits cut median readout variance with 20 Gaussian trials in two adult worm models
Reuses verified baseline-licensed panel.txt with attribution.
usesObserved site groups raised model failure clustering by 1.4 to 2.3 percent in eight partial worm maps at failure probability 0.5
Motivates site-group controls. No code, cell list or record table was reused. We test complete connection failures instead of input covariance.
usesTransferred readouts gave small median noise gains and frequent losses in two adult worm site-noise models
Reuses its verified 162-cell panel. Its within-map noise dimension test is prior context. No code or numerical results were reused.
reviewsMean single-neuron reachability loss was below 0.31 percent in eight selected worm maps with all chemical links
Prior analysis of shared-site noise in the same eight raw developmental maps. No code or result table was reused.
usesTwofold source variance bounds preserved dimension order in eight worm noise models, but larger changes reversed it
Reuses the verified panel, retained site groups and swap sampler. Tests source variance robustness of its equal-variance noise dimension finding, with new control draws and a quadratic certificate.
usesShared synaptic sites gave no clear added co-activity alignment in two adult worm maps after fixed-count controls
Reuses verified site groups, cell panel and swap functions. Adds a measured co-activity alignment test.
usesSource-count signals had lower median linear information under shared-site noise in eight worm map models
Reuses its verified cell panel and extends its signal-free noise dimension test.
reviewsBilateral neuron pairs had higher model coupling after direct link removal in two partial adult worm maps
Earlier developmental chemical-site model test informed data inspection. No data tables or code reused.