Cook chemical links gave no clear gain over degree and distance controls in ridge tests of 0.6 s calcium forecasts from 27 C. elegans

We tested whether directed chemical links select predictors of later calcium change in 27 freely moving C. elegans during copper avoidance. Main links came from the Cook graph, corrected in July 2020. The panel had 1801 target recordings for 137 native neurons. Ridge models predicted change at one imaging frame, with a median horizon of 0.600 s. Models were fitted within each animal. Gains had free signs. Controls supplied target history, behavior, shared activity, and other recorded cells with the same class label. Each wiring reference kept input and output degree for recorded neurons. Each target kept its source counts in three distance bins. Against 99 references per animal, the mean error decrease from chemical support was -0.010%. The central 95% range from animal resampling was -0.074% to +0.048%. Date-cluster resampling also crossed zero. The longer forecast at about 3 s gave no clear gain. All four planned check ranges also crossed zero. Chemical support had 0.352% more error than all recorded interclass sources, with an animal range of 0.148% to 0.541%. This limits a fixed source selection rule. It does not measure causal transmission or reject a role for chemical synapses. Limits include incomplete observations, reference anatomy from other animals, processed calcium signals, and fixed linear models. The package preserves exact native trace subsets, source versions and hashes, code, wiring control rules and hashes, forecast rules, all model losses, figures, and repeat instructions. Original anatomical tables and graph arrays remain external, with verified reconstruction steps.

connectome · calcium imaging · Caenorhabditis elegans · forecasting · chemical synapses · graph controls · copper avoidance · structure and function