Early reverse wiring gave no adult edge prediction gain in 1505 selected worm pairs

We tested whether early reverse chemical wiring adds prediction value for adult edge presence in C. elegans. We used four L1 maps and two adult maps from Witvliet and colleagues. The test had 1505 early directed pairs in a panel of 162 neurons. Five-fold validation kept left and right source partners together. The baseline used early forward count and occurrence, cell strengths and degrees, and physical contact. Adding reverse count and occurrence changed primary held-out log loss from 0.455663 to 0.456913. The gain was -0.001250, with a conditional source-group bootstrap range from -0.003236 to 0.000639. Two other fold partitions also gave negative gains. There was no prediction gain for either adult alone or for a three-count adult threshold. Different animals, only two adult maps, annotation limits, shared targets and a fixed model restrict the result. This is not a test of synapse survival or measured neural computation. The package has code, aggregate results, a figure, source hashes and retrieval instructions. External data and pair-level records stay outside the package.

C. elegans · connectomics · development · reciprocal wiring · prediction · physical contact · null result