Adult map labels raised Brier error after route count adjustment in 7,554 C. elegans atlas directions

We tested whether adult map labels improve prediction of detected neural responses. We used two adult C. elegans brain maps and the WT signal propagation atlas. All 180 shared structural neurons could act as relays. The main sample had 7,554 measured directions with no direct connection and at least three trial values. Responses were detected in 14 of 244 directions with a two-edge route only in the combined map. The count was 303 of 4,255 directions with a route in an adult map. The raw difference was -1.38 percentage points. Its descriptive cell weight range included zero. An initial model used binary pooled route presence. Adding map labels reduced pair-fold Brier error by 0.000119. A later control also used pooled relay count and its square on the log scale. Map labels then raised error from 0.05160652 to 0.05163514. Source-role and target-role holdouts also gave small error increases. All score ranges included zero. Thus, the tested models did not establish an added prediction benefit from map identity. This is an exploratory fixed-atlas test. Its ranges do not measure variation across animals. The package preserves code, input facts, source hashes, predictions, model states, weight draws, route witnesses and figures. Fresh source downloads and an offline repeat passed.

Caenorhabditis elegans · connectome · signal propagation · graph aggregation · indirect routes · relay counts · prediction · reproducible benchmark