Contact areas at shared sites reduced log area prediction error by 1.8% under a linear rule in one L3 fly circuit

We tested whether membrane contact areas vary together between different targets at one presynaptic site. The data had 540 Barnes contacts and 319 L1 plus 1,562 L3 Fritz contacts. These panels came from two fly larvae. The Fritz area key selected a different target from the redundant exported label for 131 L1 and 596 L3 rows. We checked all 1,881 area-node connector links against public skeletons and preserved both interpretations. After neuron pair means were removed, the normalized area product was -0.006, 0.112 and 0.135 in the three panels. The L3 conditional permutation p value was 0.00195. A linear rule used areas on other targets at the same site to predict a missing log area. Each slope fit excluded all contacts from the test source. L3 squared error fell by 1.80% in 998 contacts at 424 sites. The central 95% source resample range for the decrease was 0.17% to 4.18%. The two L1 point errors increased by 2.04% and 0.98%. A slope fixed at one increased L3 error by 59.79%. Thus the L3 size signal supports a small adjustment in this task. The L3 analysis was added after the first L1 test. Six sources in one L3 specimen and shared targets limit the ranges. Static areas do not identify currents, release coupling, or a change with development. The package preserves original small inputs, checked matches, provenance, code, all predictions, random scores, figures and repeat steps.

polyadic synapses · synaptic area · Drosophila · connectome · data provenance · prediction · nociception