Source groups gave opposite model orientation contrasts under 25 micrometre contact shuffles in one MICrONS mouse

We tested whether shared input segments join targets with similar model visual orientations beyond local contact controls. We scanned all 337,312,429 MICrONS v1507 synapse rows. We retained 11,615,820 input contacts to 3,461 targets, including sources with no nucleus record. The main reference shuffled contact targets within 25 micrometre cubes, target class and visual area. It kept all source and target contact counts within these groups fixed and excluded self contacts. The full-graph similarity was 0.06087 against a reference mean of 0.06197. The planned positive test gave p = 0.992. Source groups gave opposite contrasts: +0.00263 for segments with no nucleus record and -0.00252 for segments with a nucleus record. The positive subset had p = 0.001, or 0.011 after a later Holm correction over eleven secondary comparisons. All three planned reference checks kept these source-group signs. Sources with no nucleus supplied 90.87% of contacts but only 26.76% of common-input pair weight. This is an association test, not a repeat of the earlier prediction model. One mouse, fitted orientation labels, uncertain source identities and cube-based contact controls limit biological claims. The package preserves exact pair products, all reference scores, code, figures, source hashes and repeat instructions.

connectome · MICrONS · shared input · visual orientation · spatial controls · source fragments