At gain 0.5, 33 of 82 signs fixed by uncertainty bounds in a linear C. elegans chemical model opposed measured calcium signs

We tested whether unresolved chemical signs can explain errors in a linear model of C. elegans responses. We kept all 3,638 chemical edges, 20,589 contacts, and 302 neuron slots in a published graph. Each target's input weights sum to one. Uniquely predicted signs stay fixed. Complex and unpredicted weights can vary between both signs. Matrix bounds include paths of every length. At gain 0.5, the bounds fixed signs for 82 of 814 selected WT pairs with detected calcium responses and at least three finite trial values. The model agreed with 49 signs; an always-positive rule agreed with 74 on the same pairs. Changing unresolved signs cannot correct the other 33 model signs within this family. Eleven disagreements had no direct chemical edge. Across the full graph, 66 feasible sign assignments gave both response signs for 69,327 of 76,481 reachable pairs. Bounds fixed 4,888 signs. Another 2,266 pairs had no decision. All six stated gains are reported. The package supplies a short proof, the full graph, complete response atlases, reversal witnesses, figures, code, source hashes, and repeat steps. A second run matched 19 primary output files exactly. This is a test of one static chemical model. Conductance weights, electrical and peptide signals, neural state, and calcium detection can change the biological comparison. The results do not establish errors in isolated synapse labels.

connectome · Caenorhabditis elegans · structure and function · synaptic sign · interval matrix · linear response · calcium imaging · uncertainty