Data and literature evidence

The Data working group expands geographic evidence through exact publication measurements and compatible database transfers. More rows are useful only when their meaning survives extraction.

Never infer a required field from plausibility

An agent or human must not invent a coordinate, uncertainty radius, denominator, counted allele, sampling design, disease exclusion, cohort identifier, date, or source locator. Nearby place names, typical study practice, and values in a secondary compilation are not substitutes for source evidence.

If a field is not supported, preserve it as unresolved in the publication review ledger. Do not promote the row into the validated observation store.

Anchor every claim

Publication work records an immutable source record, citation, exact table/figure/page/supplement locator, verbatim population label, verbatim printed frequency, and field-level evidence. A note can explain a caveat but cannot replace a structured field.

Reuse status is checked and recorded. When no restriction is found after a documented check, use no_restriction_found; lack of an explicit license does not automatically discard valuable data. Explicit restrictions still control.

Keep database transfers source-aware

The Genome Aggregation Database (gnomAD), the 1000 Genomes Project, the Human Genome Diversity Project (HGDP), the Allele Frequency Net Database (AFND), national resources, and other datasets differ in geography, how participants were recruited, samples, and access terms. A continental genetic-analysis group is not a geographic observation. The same study group must not be counted twice when it appears in more than one source.

Follow the literature evidence curation guide and the source-specific access decisions in Issue #3.