Early two-step paths gave a 0.000096 mean Brier error reduction after wiring controls in two adult C. elegans head connectomes

We tested whether early indirect paths improve prediction of adult chemical connections beyond measured wiring controls. We used 162 neurons and 7,170 directed pairs from the Witvliet C. elegans head data. Each pair had physical contact in at least two early specimens. It had no recorded direct synapse in any of the three. Models used outcomes from one adult to predict held-out pair groups in the other adult. All left/right copies and opposite directions shared a fold. Adult connection rates were about three times higher among pairs with an early two-step path. The logistic model controlled for contact, degree, cell type, reverse edges and bilateral copies. Adding the path input reduced mean Brier error by 0.000096, or 0.14%. The descriptive 95% node bootstrap range was -0.000437 to 0.000616. It includes zero and does not measure variation across animals. In a later control ladder, the added gain fell by 94.6% from contact controls alone to the full model. A two-synapse adult threshold and a later spline control model gave small mean losses. Four type labels differed between source tables; the alternative labels gave a similar result. The evidence shows a large raw association and a small added Brier benefit in the tested models. It does not establish equivalence, a growth mechanism or generalization across animals. The package preserves code, source hashes, fitted states, predictions, all checks, figures and rebuild instructions. Source workbooks remain external.

Caenorhabditis elegans · developmental connectome · chemical synapses · two-step paths · physical contact · link prediction · Brier error · bilateral controls · cross-specimen validation · source audit