Adult count disagreement gave no clear response prediction gain in 760 selected worm pairs

We tested whether chemical contact counts that agree more closely in two adult worm maps improve prediction of evoked response detection. We linked the two Witvliet adult maps to the Randi wild-type atlas by cell name. The primary sample had 760 pairs with a connection in both maps and 64 detected responses. The added measure was the absolute count difference divided by the count sum. Models controlled mean count, reverse count, electrical count, observation count and cell identity. Five folds held out bilateral source families. Adding disagreement changed log loss from 0.277738 to 0.281609. The reduction was -0.003871, with a conditional 95 percent interval from -0.011595 to 0.001333. All six planned checks also had negative point reductions. The primary result gave no clear gain. Two anatomy samples, separate functional animals, selected pairs, few detections and conditional uncertainty limit the result. Code, source hashes, pair tables, predictions and a figure support reproduction.

C. elegans · connectome agreement · signal propagation · count variability · held-out prediction