Peptide channel count gave no clear gain in a fixed ridge test of WT minus unc-31 detection for 3,261 C. elegans atlas pairs

We tested whether possible peptide channel counts help predict a difference in neural response detection after loss of UNC-31. We joined a 302-neuron molecular map with 92 ligand-receptor channels to four anatomy exports and a fixed WT and unc-31 calcium atlas. All 301 native graph labels could act as relays. Paths stayed within each anatomy map. The main sample had 3,261 directed pairs, with at least two finite trial values in each background. The target was WT detection minus mutant detection at q < 0.05. Five test folds kept reverse directions and mirror copies together. A ridge baseline used wiring, distance, trial count and endpoint class controls. Adding log peptide channel count reduced mean squared error from 0.09304770 to 0.09304081, or 0.0074 percent. The central 95 percent node-weight range was -0.1604 to +0.1608 percent. It keeps predictions fixed and is not animal-level uncertainty. All 19 planned checks also had ranges across zero. A count term with no penalty gave a small loss. Independent source, graph and model checks passed. A clean run reproduced eight analysis output files exactly. The package preserves full molecular pair counts, complete retained native graphs, matched functional facts, code, provenance, predictions and figures. Sparse mutant trials, type-derived expression, different source animals and q-based labels limit the result. It tests a simple channel count, not the biological role of peptides or individual molecular routes.

connectome · neuropeptides · Caenorhabditis elegans · UNC-31 · signal propagation · structure and function · prediction · reproducibility