After node similarity is controlled, wiring adds at most 0.004 held-out R2 to functional coupling in worm, fly, zebrafish, mouse and human data; the anesthesia effect on mouse coupling is node-level

Question: which part of the link between wiring and functional connectivity (FC) is pair-specific, and not node or geometry structure, and where does it change with brain state? Earlier campaign work controlled nodes with one additive term per node; such a term cannot express node similarity. Data (reused): worm (40 animals, Cook wiring), fly (49 flies; rest/walk in 34; FlyWire), larval zebrafish (20 fish; swim/quiet in 12; mapZebrain), mouse fMRI (med+iso, isoflurane, awake; Knox tracer model) and human group DSI/fMRI. Method: we predicted the FC of held-out pairs in an independent replicate (other half of animals, or other hemisphere) with distance, additive node effects (M1) and a node-similarity term fitted to the other pairs of the same nodes (M2; nuclear-norm matrix completion). We compared real wiring with 30 strength- and distance-matched surrogates, split wiring into node-level and pair-specific parts, and calibrated with simulations. Results: M2 predicted most reliable FC (held-out R2 worm 0.09 to 0.75, fly 0.67 to 0.80, zebrafish 0.54 to 0.77, mouse 0.63 to 0.97; ceilings 0.88-0.99). Wiring added 0.003-0.12 after M1 (surrogates 0.000-0.002; p = 0.032, the minimum), but only -0.001 to 0.004 after M2, at most 3% of the reliable FC left. Fly, zebrafish and mouse wiring was mostly node-level (72-99% predictable from other edges). Worm wiring was mostly pair-specific and gave most of the worm gain after M1, but M2 absorbed it too (0.027 to 0.001). With synthetic wiring signals of the same size, M2 kept 6-32% (fly) and 11-25% (worm) of the gain; the real data kept -4% to 10% and 5%. The mouse state effect after M1 (med+iso 0.122, isoflurane 0.027, awake 0.022) was absent after M2 (-0.001 to 0.001) and followed a larger node-similarity gain under med+iso (0.34 vs 0.14, 0.10). Fly walk/rest and zebrafish swim/quiet did not differ. Limitations: wiring and activity from different animals; M2 is fitted to FC and can absorb wiring effects that spread to node profiles; the mouse wiring has almost no pair-specific structure, so mouse cannot test pair-specific wiring; mouse replicates share animals; reduced design (2 splits, 30 surrogates) because the host was overloaded.