Adding mean synapse volume increased mean prediction error in a fixed test of 632 C. elegans connections

We tested whether synapse size improves prediction of a calcium response after single-neuron activation. We matched 178 neurons in two adult C. elegans brain maps to a fixed signal propagation atlas. Adult 8 supplied active-zone volume and target fractions. The main sample had 632 directed chemical connections, at least three finite response values per entry, and 70 detections at q < 0.05. A ridge logistic model used counts, trial count and neuron indicators. Related directions and mirror copies shared a test fold. Adding log mean allocated volume increased mean Brier error from 0.089550 to 0.091398, or 2.06%. The central 95% network weight range for error reduction was -0.004512 to +0.000286. It includes zero and keeps fits fixed. All 16 further checks also gave mean error increases. These included source and target holdouts, size components, indirect paths and a control for correlated ridge inputs. Separate source arithmetic and a second optimizer passed. Fresh sources rebuilt the input tables. A clean run matched all 59 numerical result files. The package preserves code, complete matched graphs, source hashes, predictions, results, figures and repeat instructions. Size came from one animal, while function came from others. These results concern the tested features and atlas entries. They do not exclude a biological role for synapse size.

connectome · synapse size · active zone · Caenorhabditis elegans · signal propagation · structure and function · prediction benchmark