With detection fixed, separate removal of at most 20% of each trial list allowed 29–85 negative means in 633 C. elegans response directions

We tested how trial removal changes mean calcium response signs in a fixed C. elegans signal atlas export. The WT panel had 633 nonself directions with q < 0.05, at least five finite trial values, and no AWC endpoint. Forty-six means were negative. Removing at most floor(0.2n) values from a list could reverse 17 of the 46 negative means and 39 of 587 positive means. The other 577 signs stayed fixed under every permitted removal. With separate choices for each list and response selection fixed, the negative count could be as low as 29 or as high as 85, or 4.58–13.43% of the panel. Both endpoints were attained. One-value removal could reverse 30 signs. We checked all 633 main lists with exact rational sums. Exhaustive small-list checks and a fresh-source repeat also passed. These bounds concern fixed recorded lists. They are not confidence limits or estimates of a biological inhibitory fraction. The export has no event or animal IDs. Thus, the count bounds do not require removal of one common set of experiments. Negative calcium means are net network effects, not direct inhibitory synapse labels. The package supplies a query tool, code, derived data, endpoint witnesses, figures, source hashes and repeat instructions.

C. elegans · connectome · signal propagation · calcium imaging · response sign · trial removal · robustness · data audit