Signals from other cells reduced prediction error before and after stimulation in fixed ridge tests across 90 C. elegans worms

We tested whether signals from other cells add more prediction value after stimulation than before it. We used three matched eight-second windows from a fixed C. elegans atlas release. The same 12,004 cell measurements from 534 stimuli in 90 worms entered both periods. Source activity, cell identity, recording values, red-channel signals and anatomical wiring were supplied as controls. Each test worm was excluded from fitting and penalty selection. Signals from other cells reduced mean squared error by 4.22% before stimulation and 2.51% after it. The after-minus-before difference was -1.71 percentage points. Its descriptive 95% paired animal range was -4.87 to 1.26 points, with predictions fixed. This main test did not establish a larger gain after stimulation. A specified 40-second pulse exclusion retained 27 worms and 56 stimuli. In that panel, the gain was 1.66% before and 7.49% after stimulation. The difference range was 0.22 to 13.15 points. The smaller selected panel and unadjusted range limit this check. In 478 events, the first control window began 8 to 18 seconds after the previous pulse. Thus, this completion score alone does not isolate signal spread caused by the pulse. The windows were not randomized sham trials. Ongoing neural activity, earlier pulse effects and imaging effects remain possible sources. The package contains source code, matched window data, every prediction, fitted states, figures, source hashes and repeat instructions.

connectome · C. elegans · calcium imaging · optogenetics · prediction · spontaneous activity · matched windows · negative control