Find the hidden hubs in complex systems.

Your architecture diagram shows declared structure. IRDME measures whether the behavioral signal — real changes, interactions, flows — agrees.

When they disagree, IRDME names the discrepancy: hub_shadow, chameleon, universal_hub. Confirmed 25× across 31 pre-registered experiments.

Try the calculator →Analyse a GitHub repo →
Read the paperGitHub

31 pre-registered experiments · 25 confirmed · 4 denied with named mechanisms · arXiv:2604.23639

hub shadow

When declared and behavioral rank diverge

A hub shadow occurs when a node ranks low in the declared layer (imports, org chart, counterparty contracts) but ranks high in the behavioral layer (git co-change, transaction flow, neural activation). The gap is the structural discrepancy.

Softwarepost.php — import rank #19 → behavioral rank #4
FinanceAIG — declared tier: mid-level → derivatives hub: #1
hub_shadow— named discrepancy type in IRDME output
Results

31 pre-registered experiments

Each pre-registered before running. 25 confirmed, 4 denied with named mechanisms.

Finance
2008 Crisis · AIG
universal_hub
  • AIG declared counterparty tier: mid-level.
  • Behavioral layer (derivatives flow): rank #1.
  • Hub rank order matches crisis cascade chronology.
Software
WordPress · wp-includes
hub_shadow
  • post.php: import graph rank #19 out of 20 modules.
  • co_change layer (23,000+ commits, 20 years): rank #4.
  • r(include_graph ↔ co_change) = 0.67, p < 0.05.
r = 0.67
Neuroscience
C. elegans · connectome
universal_hub
  • Predicted AVAL, AVAR as top hubs before checking literature.
  • r(chemical_synapse ↔ gap_junction) = 0.77, p = 0.004.
  • White et al. 1986 independently identifies same neurons as primary command interneurons.
r = 0.77, p = 0.004
new

Dataset Trust Certification

PARTIAL · agreement = 55.6%

When two independent sources describe the same system, IRDME computes which nodes are structurally trusted (hub in both), which are boundary (hub in one source only — a discovery candidate or data gap), and issues a SHA-256 certificate of agreement.

Trusted nodes (5)
TP53, MDM2, ATM, BRCA1, CHK2
Hub in both BioGRID/DepMap curated v1.0 and STRING v12.0 independently.
Boundary nodes (4)
ATR, RB1, CHEK1, PCNA
Hub in one source only — data-source-dependent structural signal.
Law holds in both
r(FA, PI) = 0.749 / 0.869
Functional proximity law confirmed across independent curation pipelines.
Pre-registered before running · 5/5 hypotheses confirmedcert sha-256: ae72b231…Run your own comparison →
Tools

Topology Calculator

Describe any system as nodes and typed edges. Identifies hub nodes, cross-layer correlation (r), and labels each node by structural archetype: universal_hub, chameleon, hub_shadow, relay.

Open calculator →Browse calculator results

GitHub Audit

Paste any public GitHub URL. IRDME extracts import + co-change layers from the git log and produces a structural audit report — hub shadows, law verdict, archetype table.

Analyse a GitHub repo →
new

Project Risk

Model your project plan as Gantt + Risk layers. Identifies which tasks are schedule hubs but hidden from risk models — the failures Gantt charts miss.

Open Project Risk →
new

Composite Datasets

Merge two datasets into a single multilayer graph with an optional inter-system layer. Combine any two data sources you want to analyse as one system.

Compose datasets →

How it works

01
Define the graph
Nodes and typed edges across at least two layers. The type of each edge defines the layer: imports, co_change, chemical_synapse, derivatives.
02
Engine computes
Hub importance per layer. Cross-layer Pearson r (hub persistence). Archetype per node: universal_hub, chameleon, relay, peripheral, hub_shadow.
03
Inspect the divergence
Where behavioral rank departs from declared rank, and by how much. The output is structured JSON — no interpretation layer added.

Contact

Questions, collaborations, dataset submissions.

Open contact form →

REST API

GET /api/datasets
POST /api/run { name, config }
POST /api/datasets { name, domain, graph }
POST /api/datasets/compare { datasetIdA, datasetIdB } → trust cert
POST /api/datasets/compose { datasetIdA, datasetIdB, interLayerName, ... } → composite graph
POST /api/project-risk { graph } → risk topology
POST /api/analyze { graph } → atlas + law