Section 1 — Classifier Methodology

9-Layer Decision-Governance Engine

DiviScan GH-OS operates a 9-layer decision-governance engine. The table below discloses the exact mechanism type for every layer — distinguishing hand-coded rules, lookup databases, trained ML models, and LLM API calls.

Layer	Component	Mechanism	Description
4a	Clinical Rules Engine	Rules	26 hand-coded clinical decision rules for threshold-based alerting
4b	ClinVar Integration	Lookup	209K variant→disease mappings from NCBI ClinVar database
4c	MIMIC-IV ML Models	Trained Model	5 disease classifiers with AUC-validated performance scores
4d	ICD-10 Code System	Lookup	97K diagnostic codes for standardized disease classification
4e	Claude LLM Reasoning	API Call	Open-ended clinical reasoning across any disease via Anthropic Claude
4f	HPO Phenotype Ontology	Lookup	17,000+ phenotype→disease mappings from Human Phenotype Ontology
4g	DisGeNET	Lookup	1.1M gene→disease associations for genetic signal detection
4h	Hetionet	Lookup	2.2M biomedical edges spanning drugs, diseases, genes, and pathways
4i	PrimeKG	Lookup	8.1M biomedical edges for pathway-level disease reasoning

Transparency note: Only Layer 4c uses trained ML classifiers. All other layers operate through deterministic rules, knowledge-graph lookups, or LLM API calls. The system's primary capability is knowledge graph inference combined with LLM reasoning across 17,000+ disease phenotypes.

Section 2 — Prior Validation Signals

MIMIC-IV Classifier Performance

Five disease classifiers trained and validated on MIMIC-IV clinical data with measurable discriminative performance.

Condition	ICD-10	AUC Score	Dataset
Heart Failure	I50	0.890	MIMIC-IV
Acute Kidney Injury	N17	0.887	MIMIC-IV
Type 2 Diabetes Mellitus	E11	0.882	MIMIC-IV
Sepsis	A41	0.869	MIMIC-IV
Anemia	D64	0.815	MIMIC-IV

5 AUC-validated classifiers; all other disease associations inferred via knowledge graph + LLM reasoning.

Trial Rescue Simulation

Recoverable Trial Value · 40-Patient MIMIC-IV Simulation

$134,849,686

Recoverable Trial Value · 40-Patient MIMIC-IV Simulation

Retrospective simulation estimate — not a prospective result. Prospective validation sought.

Section 3 — Planned OSF Retrospective Re-Analysis

Independent Replication via OSF Pre-Registered Study

The next validation milestone applies DiviScan's 5 MIMIC-IV classifiers to one or more Open Science Framework (OSF)-hosted retrospective clinical datasets to produce an independent, publicly pre-registered replication signal.

Rationale: An OSF-pre-registered replication study eliminates single-dataset overfitting concerns and provides the independent signal required for Tier 1 pharma due diligence.

Scope

Run classifiers (Heart Failure I50, AKI N17, T2D E11, Sepsis A41, Anemia D64) against an OSF-registered retrospective cohort; report AUC, sensitivity/specificity, and u-metric utility scores for each classifier.

Pre-Registration

Analysis plan to be lodged on OSF prior to data access to eliminate post-hoc analytical flexibility. Pre-specified primary endpoints: AUC ≥ 0.80 on each of the 5 classifiers in the held-out cohort.

Timeline

Contingent on OSF data-access approval.

Pre-registration: Q3 2026 Results: Q4 2026

Section 4 — Tier 2 Pilot Sites

Tier 2 Health-System Pilot Targets

Four-to-six named health systems are scoped as Tier-2 pilot deployment sites for DiviScan GH-OS's HyperCore 320-classifier stack. Each card below carries a site logo, the one-line oncology/pilot angle keyed to that system's appetite, the relevant trial reference, and a direct link back to the Trial Intelligence Report engagement pitch where the pilot scope is set.

Mayo Clinic Cleveland Clinic Mass General Brigham Johns Hopkins Northwestern Memorial Vanderbilt