With axonal links fixed, endpoint counts by class and contact group fixed 10,032 of 10,129 positive axon-to-axon labels at three contacts in one fly larva

We tested whether axon-to-axon links join neurons with more shared output partners. We used the Winding et al. larval fly matrices for 2,952 cells. The main graph kept directed axonal links with at least three contacts. It had 35,815 candidate links and 10,129 positive axon-to-axon labels. Mean output-partner overlap was 0.180201 for positive labels and 0.022144 for other links. Two controls that held one endpoint fixed gave expected means of 0.167031 and 0.167261. We then kept both endpoint label counts within each cell-class pair and total-contact group. An exact graph audit showed that 10,032 positive labels, or 99.04 percent, were fixed in every valid assignment. Only 97 could change. Certified flow bounds put all feasible mean overlaps between 0.179510 and 0.180240, to the reported precision. Every valid assignment retained at least 99.39 percent of the raw overlap contrast. This narrow range follows from the stated constraints. It is not a causal explanation of wiring or proof of a neural function. The package provides code, a proof, source hashes, primary source checks, complete pair tables, figures, and integer certificates. Fewer than eight percent of positive labels could change in each listed check. The study is exploratory and applies to one fixed graph. It does not estimate a probability over valid assignments.

connectome · Drosophila larva · axon-to-axon connections · null model · fixed margins · forced labels · minimum-cost flow · reproducibility