A model with no internal axon-to-dendrite transfer lost two-step reachability for 46,259 pairs in one larval fly brain graph

We tested how transfer inside neurons changes short paths in one published larval Drosophila brain graph. We used 2,952 neurons and all four chemical synapse types from Winding et al. Data S1. The main model allowed dendrite-to-axon internal transfer but blocked axon-to-dendrite internal transfer. A path could start and end in either arbor. At least one contact was required for each edge. A whole-neuron graph reached 1,823,803 ordered pairs within two synaptic steps. The model reached 1,777,544 of them. The difference was 46,259 pairs, or 2.54 percent. Most short reachability remained. With at least five contacts per type edge, the difference fell to 447 of 173,101 reached pairs, or 0.26 percent. At any finite length, both unfiltered graphs reached the same 8,291,500 pairs. The model lengthened the shortest route for 343,715 pairs, or 4.15 percent. Allowing reverse transfer at 110 listed local neurons reduced the two-step difference to 43,845 pairs. The axon-to-dendrite-only graph had no difference. The package supplies an exact formula, a path query tool, compact inputs, source hashes, all primary distances, controls, figures, and repeat instructions. Primary supplement tables matched the pinned mirror byte for byte. These are conditional results for one annotation file. The one-way rule is a limiting model assumption. It is not a measured property of the animal, and the results do not establish a loss of neural communication.

connectome · larval Drosophila · synaptic polarity · axon and dendrite · shortest paths · graph aggregation