FPL at 139,000 Neurons: Structural Hubs Persist in the Largest Brain Connectome Ever Mapped

The Largest Complete Connectome Test of FPL

The FlyWire Princeton connectome (Dorkenwald et al. 2022) contains 139,255 neurons and 5.3 million directed synaptic connections -- the largest complete animal brain ever mapped. We pre-registered five hypotheses (DISC_F14_FLYWIRE_v1, hash ea427603, commit 787a4de, 2026-06-14) before running any analysis.

H1 CONFIRMED: r(d1, d2) = +0.819 at 129,117 connected neurons. This is the strongest FPL correlation ever measured in a connectome.

FPL Strengthens with Brain Size (3-Point Trend)

Across three independently pre-registered connectome experiments spanning a 428x scale range:
• C. elegans (302 neurons, F12): r = +0.317
• Larval Drosophila (2,952 neurons, F13): r = +0.663
• Adult Drosophila (129,117 neurons, F14): r = +0.819

FPL does not merely hold at scale -- it strengthens. This is a post-hoc observation from three data points, not yet pre-registered as a scaling law. A fourth species (zebrafish or mouse cortex) will test whether this is a genuine scaling relationship.

The optic lobe alone (87,859 neurons) produces r = +0.848 -- stronger than the whole-brain result. Retinotopic regularity amplifies hub persistence: the columnar grid structure of the medulla creates correlated fan-out and fan-in at every spatial position.

Finding: Global Inhibition Is the Structural Backbone

We pre-registered that ACH (acetylcholine, excitatory) neurons would be the top d1 hubs. The actual result was the opposite.

The top hub in adult Drosophila is the APL (All-Projection Lateral) neuron: a single GABAergic inhibitory neuron per hemisphere with d1 = 14,000 and d2 = 14,786. It contacts every Kenyon cell in the mushroom body -- the brain's learning and memory center.

GABA neurons have mean d1 = 71.15 versus ACH mean d1 = 40.12 (t = -31.68, p < 10^-218). The hub structure of the adult fly brain is dominated by global inhibitory broadcast neurons, not excitatory projection neurons. Structural backbone = inhibition at scale.

Finding: Dopamine Neurons Are Universal Hubs

Despite comprising only 461 neurons (0.3% of the network), DA (dopaminergic) neurons have the highest mean d2 of any neurotransmitter type: 174.77 versus ACH 37.15.

Dopaminergic mushroom body neurons receive convergent input from the widest variety of sources and broadcast reward and value signals broadly. They are the reward-signal broadcast system: structurally, the most efficient universal hubs in the brain.

Finding: Binary Connectivity > Synapse Count Weighting

We pre-registered that synapse-count weighting would strengthen hub persistence (H2). Result: r(weighted) = +0.803 < r(binary) = +0.819. Not confirmed.

Binary contact diversity -- how many distinct neurons you connect to -- is a more fundamental structural signal than total synapse count. A neuron that contacts 14,000 distinct targets is a stronger hub than one that makes 14,000 synapses on fewer targets.

Hub Regime: Generalist at Brain Scale

In npm software, hub overlap between top-1% d1 and top-1% d2 is near zero (consumer vs infrastructure polarity). In adult Drosophila, 47.9% of top-1% d1 neurons are also top-1% d2. The same neurons both receive from many sources and project broadly.

This is mechanistically necessary: the APL neuron must integrate inputs from all 14,000 Kenyon cells AND inhibit all 14,000 simultaneously. The generalist hub regime is a structural consequence of global modulatory architecture.

Pre-Registered Before Analysis

All five hypotheses were written, hashed, and committed to the public preregistrations repository before any degree distribution was computed. The GABA > ACH result was a genuine pre-registered null -- the scientific workflow was preserved.

Pre-registration: DISC_F14_FLYWIRE_v1, hash ea427603..., commit 787a4de, 2026-06-14.

---

IRDME is a pre-registration-first framework for structural multilayer network analysis.