How SaudiVarKB is built
A literature-derived resource designed around traceable evidence, reproducible scope rules, and conservative release criteria.
Literature scope
The harvested corpus is biomedical literature with Saudi institutional linkage: PubMed matched the Saudi search terms in any field, including author affiliation, so most records do not study a Saudi population. A reproducible rule identifies 3,447 Saudi human-genetics articles; in an independent review of 100 randomly sampled core articles, 67% (95% CI 57.3 to 75.4) met predefined Saudi human-genetics eligibility criteria. Searches use this core subset by default, and the full set remains available.
Extraction pipeline
- Publication metadata and abstracts are normalized into stable article records.
- Genes, sequence variants, phenotypes, and population context are extracted.
- Every mention preserves its source sentence, extraction method, and confidence.
- Quality rules retain supported mentions and exclude ambiguous candidates.
Quality and limitations
The release exposes 69,633 retained mentions and withholds 23,825 excluded candidates. Automated extraction can still contain errors. Presence does not establish pathogenicity, prevalence, ancestry, or clinical validity; the source publication remains authoritative.