With 100 micrometre source blocks, a 10 micrometre path test on 274 MICrONS dendrites gave model orientation excess 0.068 with a target resample range of -0.032 to 0.170

We tested whether inputs from different neurons have similar model visual orientations when they are close along a dendrite skeleton. We reused an extract built from all 337,312,429 released MICrONS v1507 filtered synapse rows. We matched public static v1300 SWCs by exact root ID. This gave 1,275 targets and 16,465 contacts. A two micrometre projection limit retained 13,583 contacts. The primary test used 274 targets, 645 near source pairs at path distance at most 10 micrometres, and 1,579 reference pairs at greater than 20 to 50 micrometres. Source function records were shuffled within area, cell type, and 100 micrometre soma cubes. The orientation score was 0.07256. Its exact reference mean was 0.00446. The excess was 0.06810, with a target resample range of -0.03244 to 0.16979 and an upper-tail probability of 0.0625 from 1,999 draws. A 50 micrometre source reference gave probability 0.017, but most other checks were less clear. A four micrometre projection limit reduced the excess to 0.02714 on the same 274 targets. In 272 matched targets, 91 of 729 source pairs near by straight distance had no path within 10 micrometres. Including contacts from one source raised the path contact score from 0.07081 to 0.21010. The evidence for grouping of distinct inputs remained sensitive to the tested choices. The package preserves code, exact source records, compact distance data, figures, all results, and independent graph checks. One mouse, sparse source coverage, model tuning, and nearest-edge projection limit the conclusion.

MICrONS · dendrite · connectome · orientation tuning · synaptic input · skeleton distance · source identity · spatial control