Early chemical edge presence gave no clear response prediction gain in 1316 selected worm pairs

We tested whether chemical edge presence in three early larval maps adds adult response detection prediction value beyond adult synapse count. We linked eight Witvliet maps to the Randi wild-type atlas by exact cell names. The sample had 1316 directed pairs and 109 detected responses. Five folds held out stimulated bilateral cell families. Adding early-map occupancy changed log loss from 0.266394 to 0.266318. The reduction was 0.000076, with a conditional 95 percent crossed-family bootstrap interval from -0.005544 to 0.005052. Seven planned sensitivity checks and a post-list counting check also had intervals that included zero. Early occupancy increased with adult count. These data gave no clear added prediction value under the tested model. Small gains remain possible. Separate animals, few anatomy samples, selected cells, pooled response tests and unequal count overlap limit the result. Code, pair data, hashes and a figure support reproduction.

C. elegans · development · connectome · signal propagation · held-out families · null result