Electrical weights gave no clear gain for new pairs of neuron classes in a ridge test of calcium coactivity in 27 C. elegans

We tested whether reference electrical weights help predict calcium coactivity between different neuron classes. The data came from 27 freely moving C. elegans in arenas with a copper boundary. The primary panel had 22,450 observations of 1,809 class pairs. We removed a fitted span of measured behavior and boundary-distance features. Each test pair was excluded from fitting in every animal. Controls supplied class identity, soma distance, chemical strength and shared chemical profiles. With separate ridge penalties, electrical weights increased mean squared error by 1.75%. With the control penalty used for both models, they reduced error by 1.66%. The fixed-prediction 95% ranges from resampling animals and classes were -12.43% to +4.12% and -2.71% to +5.64%. Ranges from 199 refits also crossed zero. A second adult map gave a positive point gain with an uncertain class range. The evidence gives no clear gain across class identities for these fixed prediction rules. Both electrical maps are incomplete. The results do not estimate junction conductance or establish absence of an electrical effect. The package preserves source records, derived data, code, models, test predictions, figures and repeat instructions.

C. elegans · connectome · calcium imaging · electrical synapses · behavior adjustment · prediction · neuron classes · reproducible data