Two receptor sign terms raised error by 0.14–0.17% in two models of behavior-adjusted calcium coactivity from 21 C. elegans recordings

We tested whether predicted chemical synapse signs improve calcium coactivity prediction. We used all 302 neuron slots and all positive nonself links in the corrected Cook hermaphrodite graph. Transmitter and receptor expression rebuilt all 3,638 original Fenyves sign labels. Unique signs covered 48.7% of Cook chemical links and 57.6% of chemical weight. The activity benchmark had 75,602 pair entries from 21 freely moving worms. Every test row excluded its animal and both endpoint cell-class groups from training. Controls included unsigned wiring, electrical links, sign coverage, cell attributes and a transmitter-only rule. Adding direct-link and shared-input receptor sign terms raised mean test error by 0.141% in the planned least-squares model. A later mean-reference check showed that this model was weak. Nested ridge selection then reduced control error by 5.52% against the mean, but the two sign terms raised error by 0.166%. Conditional animal resample ranges for the two sign gains were -0.252 to -0.030% and -0.245 to -0.088%. A direct sign term alone gave a small favorable fixed-test result. Longer behavior controls also changed the joint result. The evidence limits these specific prediction rules. It does not establish incorrect synaptic signs or no biological role for inhibition. Missing signs, indirect network effects, slow calcium signals and unmatched anatomical animals limit the test. The package contains code, full-node sign coverage, exact covariates, scores, model choices, predictions, figures, source hashes and repeat instructions.

C. elegans · connectome · structure and function · synaptic sign · receptor expression · calcium coactivity · common input · nested validation · ridge regression