With known contacts, filters and capacitances, 55 sine patterns gave unique weights from exact responses in a calibrated, fully observed 302-cell worm model

We studied a linear worm model with 302 fully observed cells. The contact map, cell capacitances and chemical filters were known. A row-count proof showed that 55 complex sine patterns were sufficient to fix all weights and cell leaks. The map had 3,671 chemical directions and 1,091 electrical pairs. We tested one stable synthetic network with 20 paired pattern and noise draws. Each design had the same total squared current. Noise was scaled by the root mean square (RMS) voltage of the full single-cell response. At complex noise RMS of 0.1% of this reference, 55 patterns gave 6.22% median chemical-weight error. The central 80% range was 3.78–41.76%. With 63 patterns, the median error was 2.27% and the range was 2.20–2.36%. A 302-experiment single-cell reference gave 1.84%. After electrical removal, 63 patterns gave 2.24% median error in DC responses between different cells. Complete-matrix error was 0.0497% because large self responses hid much of the propagation error. We added this control after the first calculation. No animal activity data were fitted. The bound is sufficient for this row method. It is not a minimum for every reciprocal estimator. Full calibration, known kinetics and full state access limit use. The package has the proof, code, source audit, contact tables, results, figures and repeat steps.

connectome · system identification · patterned stimulation · electrical synapses · chemical kinetics · C. elegans · noise · reproducible methods