GaiaLab Trust & Transparency

Live Trust Scoreboard

These metrics reflect the most recent analysis run stored in this browser. Evidence Density = signal volume relative to gene inputs. Contradiction Index = fraction of papers with conflicting signals (target <15%). Citation Coverage = claims backed by ≥1 PubMed citation, post-polarity gating. Grounded Evidence = pathways & strategies with confirmed supporting polarity — supplementary topic themes excluded from denominator. Quality Score = weighted sum of paper count, insight count, and full-text bonus, scaled by a grounding multiplier (0.5–1.0). Hover each metric for exact definitions. Run a new analysis to refresh.

Evidence Density (EDS) ?

Signal volume relative to inputs. Target: >60%

Contradiction Index (CI) ?

Conflicting evidence ratio. Target: <15%

Citation Coverage ?

Claims backed by PubMed citations. Target: >70%

Quality Score ?

Composite confidence. Target: ≥80/100

Cite-F1 (ALCE) ?

Claim-level citation quality. Target: >60%

Quality Score — Run History

No run history yet. Complete an analysis to start tracking.

No analysis data available. Run an analysis from the dashboard to populate.

How to Read These Signals

Trust signals update after each analysis run. Use Evidence Density and Citation Coverage to gauge data richness. Monitor the Contradiction Index to identify areas requiring manual review. The Quality Score provides a single composite measure for triage decisions.

Latest Run Detail

Detailed breakdown of the most recent analysis. Data is stored locally in this browser session.

No analysis data available. Run an analysis from the dashboard to see results here.

Pipeline Overview

A six-stage pipeline that keeps evidence provenance, scoring, and replayability transparent at every step.

1. Input Normalization

2. Multi-source Retrieval

3. Cross-validation

4. Consensus Scoring

5. Evidence Ledger

6. Snapshot Archive

Evidence Ledger Format

Each claim is tracked with source identifiers, confidence tier, and constraint notes.

Claim	Evidence IDs	Confidence	Notes
EGFR is implicated in NSCLC	PMID:123456; OpenTargets:OTAR1234	High	Strong agreement across sources
KRAS drives MAPK activation	PMID:987654; KEGG:hsa04010	High	Consensus across pathway databases

Confidence Scoring Policy

Confidence is consensus-weighted across independent sources. The formula applies diminishing returns, contradiction penalties, and a hard cap at 0.95 to prevent false certainty.

Scoring formula

agreement_bonus = 0.12 * (1 - exp(-0.9 * A))
contradiction_penalty = 0.25 * CI
final = clamp(base + bonus - penalty, 0, 0.95)

A = number of independent source domains in agreement
CI = contradiction index for this claim

Contradiction Handling

When sources disagree, GaiaLab does not average away conflict. Contradictions are surfaced with supporting and refuting evidence for manual adjudication.

Example: Conflicting evidence across cohorts
Severity: Medium
Support: PMID:112233
Refute: PMID:445566

Reproducible Snapshots

Every analysis produces a replayable snapshot containing inputs, source versions, scoring policy, model configuration, evidence ledger, and visualization state.

View latest snapshot JSON Load sample inputs

Snapshots are also accessible at /analysis/{id} after each run. The analysis ID appears in every API response under analysisId.

Prediction Calibration

GaiaLab prospectively records drug repurposing predictions and checks them against ClinicalTrials.gov outcomes. A well-calibrated system's 80% confidence predictions should be correct ~80% of the time.

Loading calibration data…

Safety & Scope

Research and exploratory use only.
No dosing recommendations or clinical decision support.
No patient-specific diagnostic or treatment advice.
Low-evidence findings are labeled as hypotheses with suggested validation paths.