A flag for two-edge chemical routes in an adult map increased Brier error in a fixed model of calcium response detection for 3,293 C. elegans atlas pairs

We tested whether a short chemical route contained in an adult brain map helps predict a detected calcium response. We used two adult C. elegans connectomes and the Randi signal propagation atlas. All 3,293 primary pairs had a shortest distance of two edges in the pooled graph, at least three usable trials, and no direct left/right partner relation. Of these pairs, 3,051 had a two-edge route in at least one adult. Their response detection rate was 6.29%, against 5.79% in the other 242 pairs. Adding the route flag to a fixed logistic model increased Brier error for entries held out from fitting from 0.056329 to 0.056481. Each of eight planned extra checks also gave a higher Brier error. A later scope check found alternative short routes in 164 of the 242 pairs when electrical edges and older anatomy exports were considered. The package includes graph calculations on up to 298 atlas labels, derived pair data, fixed folds, code, all predictions, model states, figures, source hashes, and repeat steps. Independent calculations checked distances, response labels, and predictions. These results apply to fixed atlas entries and models. The anatomy and response animals differ. The test does not show that within-animal routes have no functional value or that the compared response rates are equal.

connectome · Caenorhabditis elegans · signal propagation · calcium response · graph pooling · anatomical paths · structure and function · reproducible benchmark