Pre-Registration: GaiaLab Drug-Repurposing Prospective Prediction Set

GaiaLab is an AI biological-intelligence platform that ranks drug–disease repurposing candidates. This record pre-registers a prospective test of its forward predictive value. Most of GaiaLab's ~20,000 logged drug–disease predictions have been matched to ClinicalTrials.gov trials that already existed at prediction time (retrospective concordance). That cannot demonstrate forward predictive value. This pre-registration freezes a set of high-confidence predictions that currently have no registered trial, and commits — before any outcome is known — to a falsifiable hypothesis about their future trial-registration rate. Frozen hypothesis. Among GaiaLab drug–disease predictions with no ClinicalTrials.gov trial at lock time (2026-07-17), the Tier II+ set (proprietary repurposing score ≥ 50) will register new interventional trials by 2028-01-17 at a higher rate than a matched Tier III control set (score < 50). Frozen cohort. 357 prediction-arm pairs (score ≥ 50, no trial at lock) vs 148 matched Tier III control pairs. Non-therapeutic entries (reagents/supplements) were excluded before sealing. Cohort fingerprint (SHA-256): daaed957d2d38171b6a626b11d2a6df48a6262165fc55c3e685b1c3a64bfb96b This one-way hash timestamps the exact frozen prediction set. At readout, anyone given the frozen cohort can recompute it and confirm the set was fixed in advance — no backdating possible. The GaiaLab platform code is proprietary and is not disclosed; only the claim and this fingerprint are. Endpoint & statistic (pre-declared). Outcome = a ClinicalTrials.gov interventional trial for that drug in that disease first-posted after 2026-07-17 and on/before 2028-01-17. Metric = prediction-arm hit rate vs control-arm hit rate; Fisher's exact two-tailed test. Supported iff p < 0.05 and hit-rate ratio > 1; refuted iff p < 0.05 and ratio < 1; else inconclusive. Commitments. The cohort is immutable (verified by re-hashing at readout). Predictions are evaluated exactly as scored on 2026-07-17 — no re-scoring, no peeking. The readout, including a null or refuting result, will be published on/after the evaluation date. A refutation is a valid scientific outcome. Limitation. New-trial registration is a proxy for "worth testing," not proof of efficacy. Drug popularity confounds trial rates; the matched control absorbs the baseline but residual confounding remains.