With a common decoder and five observed cells, position and cell groups cut calcium response error by 3.2% versus random choice in 83 C. elegans worms

We tested whether electrical wiring helps choose which neurons to observe. We used 506 single-neuron stimulation events in 83 C. elegans worms. Each split kept a common hidden query set and a separate candidate set. Three rules chose five candidates. All rules used the same decoder, which had access to electrical covariance. Test animals stayed out of fitting and decoder choice. Selection from soma position, bilateral class and pharyngeal membership decreased mean squared error by 3.240% versus random choice. Its conditional 95% animal resample range was 2.240% to 4.214%. A second split gave a 2.800% decrease. Adding an electrical graph to selection gave a 0.077% further decrease, with a range of -0.154% to +0.315%. The two anatomical rules shared 94.2% of their chosen cells. Even under the assumed covariance, the expected added graph gain was only 0.029%. Thus, this test had little contrast between the anatomical rules. A red-channel check also gave a small selection gain. The results support position and cell-group choice in this fixed response completion task. They do not identify a neural cause or establish that electrical wiring lacks useful information. The source panel was selected, the task used retrospective masks, and resample ranges keep fitted models fixed. The package preserves code, data facts, source records, model states, masks, predictions, figures and repeat commands.

connectome · neuron selection · experimental design · calcium imaging · C. elegans · electrical synapses · response completion