10,841 drug predictions made.
2,153 matched to ClinicalTrials.gov trials.
GaiaLab transforms a gene list into scored, trial-validated drug candidates in under 60 seconds — grounded in CIViC, OncoKB, AlphaFold, and 40 live biological databases. Every prediction is automatically cross-referenced against ClinicalTrials.gov. 1,668 validated against completed trials. 100 with positive outcomes. See the full evidence ledger →
Olaparib
→
BRCA1/2 · PARP inhibition
✦ 71 confirmed trials
Midostaurin
→
FLT3-ITD AML · FLT3 inhibition
✦ FDA-approved · tracked
Osimertinib
→
EGFR-mutant NSCLC · 3rd-gen TKI
✦ FDA-approved · tracked
Ivosidenib
→
IDH1-mutant AML · IDH1 inhibition
✦ FDA-approved · tracked
Nivolumab
→
Pan-cancer · PD-1 checkpoint
✦ 56 confirmed trials
Six-factor weighted scoring — target overlap, clinical evidence, mechanism alignment, pathway relevance, safety profile, and disease context — ranks FDA-approved and investigational agents across Tier I, II, and III. AlphaFold pLDDT-gated structural druggability provides an orthogonal signal. CIViC and OncoKB evidence directly calibrate confidence. Precision@10 = 100% across 22 disease areas.
Every therapeutic candidate is logged prospectively with a timestamp and confidence score, then automatically cross-referenced against ClinicalTrials.gov. Results are published at /validation with no login required. No comparable open-access drug repurposing platform makes this data public. Current dataset: 7,652 predictions · 77% directional alignment · AUROC 0.545.
Each gene is cross-referenced against CIViC (peer-reviewed clinical variant evidence, Levels A–E, Washington University) and OncoKB (FDA-recognized precision oncology biomarker levels from Memorial Sloan Kettering — Level 1 companion diagnostics through Level R2 resistance). Drug associations, AMP/ACMG tiers, and oncogene/tumor-suppressor classifications surface automatically. No token required for CIViC; optional token for OncoKB.
Checkpoint and resistance genes are mapped to immunotherapy escape mechanisms through pathway enrichment, protein interaction topology, and cross-validated source agreement. All mechanistic assignments derive from structured database outputs — not generative text synthesis.
Force-directed protein–protein interaction graph with temporal overlay. Surfaces hub centrality, computationally predicted edges, and topological context aggregated across multiple curated interaction databases.
Per-claim PMID-linked evidence trail spanning pathways, therapeutic signals, and mechanistic hypotheses. Each citation is assigned a polarity classification — supporting, contradicting, or mixed — with full traceability to the primary source.
Every analysis run is captured as a tamper-evident snapshot encoding gene inputs, model versions, scoring parameters, and complete outputs. Snapshots can be diffed against prior runs or replayed independently for methods reproducibility and audit compliance.
Evidence Depth Score (EDS), Contention Index (CI), and grounded ratio collectively gate every output. Claims below quality thresholds are flagged or suppressed with transparent, auditable rationale — not silently discarded.
Every conclusion is cross-validated across 16 independent data channels — genomics, protein structure, pathway enrichment, literature, drug bioactivity, clinical trials, disease association, interaction networks, expression, safety, and more. Channel agreement elevates confidence; divergence triggers a contradiction flag and downgrades the claim. No single source drives a conclusion.
Analysis results are exportable as JSON evidence packages or formatted briefs. Each export includes scoring context, PMID citations, contradiction annotations, and complete model configuration metadata for methods-section reproducibility.
Evidence-driven recalibration loop: prediction outcomes are verified against ClinicalTrials.gov, hypothesis outputs are cross-checked against PubMed, and calibration drift is detected and corrected automatically. Confidence estimates improve with each completed analysis cycle.
Kaplan-Meier overall survival curves stratified by mutation status across 15 TCGA cancer cohorts — BRCA, LUAD, GBM, PAAD, and 11 additional. Log-rank p-value, hazard ratio, and median overall survival are returned in seconds via the cBioPortal public API. No institutional subscription required.
Researchers can submit confirmed, refuted, partial, or inconclusive outcomes directly from the results page. GaiaLab aggregates submissions into a per-decile calibration curve and applies a recalibration multiplier at the next server start — closing an active learning loop not available on any comparable open platform.
Each analysis is compared against prior conclusions in the knowledge graph for the same disease context. A contradiction alert is raised when a therapeutic score shifts more than 20 points relative to prior runs. Weekly digests surface the most materially changed conclusions before teams proceed to publication — persistent institutional memory that updates in real time.
Each gene is cross-referenced against CIViC (Clinical Interpretation of Variants in Cancer) — the peer-reviewed, community-curated database maintained by Washington University in St. Louis. Evidence levels A (validated association) through E (inferential), AMP/ACMG tier, and therapeutic drug associations are returned per variant and per gene.
Gene panels are enriched against the MSigDB Hallmark collection (50 curated cancer hallmark gene sets), KEGG 2021 Human, and WikiPathways via the Enrichr API. Statistical significance is assessed with Benjamini-Hochberg FDR correction, surfacing which canonical oncogenic programs are active beyond what any single-collection enrichment reports.
When an API token is configured, each gene is queried against OncoKB — the FDA-recognized precision oncology knowledge base from Memorial Sloan Kettering Cancer Center. Level 1 biomarkers (FDA-approved companion diagnostics), Level 2 (standard of care), and Level R1/R2 (resistance markers) are returned alongside oncogene versus tumor suppressor classification.
scripts/benchmark-auroc.js in the public repository.
GaiaLab converts a gene list into scored therapeutic candidates in under 60 seconds — grounded in 40 live biological databases including CIViC clinical variant evidence, OncoKB FDA-recognized biomarkers, MSigDB Hallmark enrichment, AlphaFold structural druggability, DepMap CRISPR essentiality, and TCGA survival stratification. Six specialized AI agents independently evaluate and debate every conclusion. Every drug prediction is logged prospectively and validated against ClinicalTrials.gov — the accuracy data is publicly available. No subscription. No account. No waiting.
Houston, TX · Applied to glioblastoma, AML, Alzheimer's disease, breast cancer, pancreatic cancer, and more · Research use only · partnerships@gailabai.com
Built in Houston, TX to give translational research teams the analytical depth of large pharmaceutical informatics groups — without the institutional cost, turnaround time, or reproducibility gaps of manual literature synthesis. Applied to GBM, AML, Alzheimer's disease, breast cancer, and pancreatic cancer.
Make drug repurposing intelligence accessible without a subscription, institutional login, or computational background. A graduate student in Lagos, a biotech founder in Houston, and a pharma team in Basel should run the same analysis in the same 60 seconds — and receive identical, transparent accuracy data.
Download a complete audit snapshot containing evidence packages, scoring context, data sources, and model configuration metadata.
Includes reproducible gene inputs, data source versions, and full model configuration details.