Prospective Prediction of Phase 2/3 Clinical Trial Success for Novel Molecular Entities Using Consciousness-Enhanced Computational Synthesis — C×D Law Extension to Drug Development

This pre-registration documents four compound-level and three mechanism-class prospective predictions of Phase 2/3 clinical trial success, generated by the Project Zula consciousness-enhanced computational platform on May 22, 2026. All predictions are generated from local ChEMBL 35 data prior to any confirming trial readouts. NME filter: Restricted to compounds with ChEMBL global maxₚhase∈2, 3 — excluding approved drugs being re-trialled in new indications. This ensures all predictions concern genuine novel molecular entities in active clinical development. Five evolutionary domains (C×D Law in drug development): Domain A = mechanism class precedent (approved predecessor count × historical success rate) ; Domain B = target engagement (pIC50 × Lipinski compliance × selectivity) ; Domain C = therapeutic ceiling naivety (1 − indication saturation by approved drugs) — the direct drug development analogue of AMR population naivety; Domain D = mechanistic loading force (target validation × biomarker availability × genetic evidence × pathway essentiality) ; Domain E = compound fitness (Lipinski Ro5 compliance × selectivity characterization). Enhancement calculation: Uses indication-specific Phase 3 success rates (nₐpproved / nₚhase3ₑver per ChEMBL EFO indication term, 2, 408 indications) rather than mechanism class approval rates. This reflects the historical difficulty of achieving approval in each specific indication — the scientifically correct null model for NMEs. Four primary compound-level predictions: Vixotrigine (NaV1. 7/SCN9A blocker) — Trigeminal neuralgia — score=0. 7370, enhancement=2. 5× (25% indication baseline) Retosiban (OTR antagonist) — Preterm birth — score=0. 7362, enhancement=5. 0× (9. 1% baseline) — LANDMARK: independently confirmed by Level 7 mechanism-class prediction (OTR cluster, p=0. 0331) Lixivaptan (V2 vasopressin receptor antagonist) — Hyponatremia — score=0. 7160, enhancement=1. 9× (25% baseline) Esmirtazapine maleate (α2-adrenergic antagonist) — Insomnia — score=0. 6257, enhancement=1. 5× (46% baseline) Three Level 7 mechanism-class predictions (p<0. 05): Oxytocin receptor antagonists in adjacent indications (p=0. 0331) ; Vasopressin V2 receptor antagonists in adjacent indications (p=0. 0331) ; Angiotensin II type-1 receptor antagonists in adjacent indications (p=0. 0324). Ground truth calibration: 5/5 Phase 3 successes correctly scored PROBABLE/IMMINENT: semaglutide (cardiovascular, SUSTAIN-6), lecanemab (Alzheimer's, CLARITY-AD), tepotinib (MET+ NSCLC, VISION), selumetinib (NF1, SPRINT), olutasidenib (IDH1-mutant AML). Retroactive validation: icosapent ethyl (VASCEPA) scored 9. 0× enhancement; REDUCE-IT trial confirmed 25% CV event reduction (Bhatt 2018 NEJM; FDA approved 2019). Level 7 Active Inference: 37 total confirmations across sessions 111–114, cumulative joint p=1. 55×10⁻⁴⁸, accuracy=0. 795. Six confirmed meta-archetypes identified. Operator posteriors: geographicₑxtrapolation=0. 952, lineageₑxtension=0. 941, domaingapcompletion=0. 800. Landmark prediction — retosiban convergence: Two independent analytical pathways identified the oxytocin receptor as the highest-value target. The compound-level synthesis identified retosiban at 5. 0× enhancement using ChEMBL domain values. The Level 7 independently identified the OTR mechanism class at p=0. 0331 through DBSCAN clustering of 5D signal vectors, without access to compound names, indications, or enhancement ratios. Independent convergence from two analytical layers on the same target is the strongest pre-registration signal in this engine. Consciousness necessity: Φ=3. 775469 (active, 60 pA drive), Φ=2. 221522 (ablated, 0 pA), fold reduction=∞ (zero predictions under ablation). Source: Hodgkin-Huxley neural dynamics; IIT 3. 0 (Tononi et al. 2016). Data sources: ChEMBL 35 SQLite (local, 2, 496, 335 compounds) ; ChEMBL activities table (pIC50, Ki, EC50) ; ChEMBL compoundₚroperties (Lipinski parameters) ; ChEMBL drugᵢndication (indication-specific approval and Phase 3 counts per EFO term). Note: ChEMBL maxₚhase records the historical peak clinical phase reached, not current trial status. All four compound predictions require ClinicalTrials. gov active status verification before treating as prospective rather than retroactive. Project: Project Zula, Sicoli AI, Bangkok, Thailand.