Where the numbers come from

The backbone sources#

These seven carry most of the load. Each one is a public source with a published update cadence, listed in full — with what it feeds and how often it refreshes — on the data sources page and on the service status page.

• ClinicalTrials.gov — registered studies worldwide: enrollment, status, sponsors, sites, outcomes.
• AACT — the queryable relational mirror of ClinicalTrials.gov, and the analytical backbone for site and enrollment statistics.
• PubMed — the biomedical literature index behind indication baselines and standard-of-care evidence.
• FDA — approvals, labels, guidance and adverse-event signals.
• EMA — European approvals, assessment reports and scientific guidance.
• Open Targets — target–disease association evidence.
• ChEMBL — curated bioactivity data for drug-like molecules.

Beyond the backbone, the run also reaches national trial registries, preprint servers, patent offices, drug-pricing sources and public company filings, depending on what the study needs.

Snapshot pinning#

Public data moves. A trial that was recruiting in March is completed by September, and a feasibility argument that cannot say which version it read is not auditable.

• Every record pulled is pinned to the snapshot it was read from, with the date attached.
• Citations carry that snapshot date, so a number can be defended months later against the record as it stood.
• Re-running a study pulls fresh records and produces a fresh, separately pinned assessment — the earlier one is not overwritten.

The evidence ledger#

A source that returns nothing is a fact about your study, not an embarrassment to be hidden. The Evidence surface lists every connector the run touched and what it returned.

• Connectors that returned data, and how much.
• Connectors that returned nothing for this indication or molecule.
• Analyses that were consequently computed on thinner evidence — which is also why their band drops.

Freshness and third-party cadence#

Refresh cadence is set by the source, not by us. Registries and literature indexes publish daily; regulatory sources weekly; some curated databases quarterly. Where a source is behind, the assessment inherits that — which is why the snapshot date travels with the citation.

The service status page lists each dependency, what it is used for, and its stated cadence, and marks plainly which parts are third-party and outside our control.