Methods & evidence policy

Discovery is automated.
Trust is reviewed.

ACBDT is a research and education platform. It does not diagnose, recommend treatment, or convert literature matches into clinical claims.

01

Discover candidates

Versioned country × cancer queries use the Europe PMC REST API. Search results begin as UNSCREENED candidates.

02

Screen relevance

A curator checks whether the paper is relevant, whether a country is merely mentioned, and whether full text is required.

03

Structure the finding

Cancer, gene, variant, protein, assay, cohort, clinical role, and numerical finding remain separate entities.

04

Review and approve

Source excerpts, reviewer identity, rationale, timestamp, and publication status travel with every approved finding.

Scientific semantics

The distinctions the platform will not collapse.

Mention
A term found in searchable text. It is not evidence of a cohort or a finding.
Candidate
A publication that may be relevant and needs scientific screening.
Cohort verified
Recruitment or sample origin has been confirmed from an appropriate source passage.
Approved finding
A structured claim with assay, cohort, source evidence, and review decision attached.

Plain-language glossary

Words that should stay precise.

Gene

A DNA unit that can encode a product. Example: ERBB2.

Variant

A specific change in a DNA sequence. It is not interchangeable with a gene.

Protein

A biological product that may be measured. Example: HER2 protein.

Assay

The laboratory method used to measure a target, including platform and cutoff.

Biomarker claim

A context-specific relationship between a measured target and a clinical role.

Clinical role

Diagnostic, prognostic, predictive, risk, monitoring, response, or safety.

Pilot source

Europe PMC via a reproducible API workflow.

The script records query text, result identifiers, timestamps, candidate status, and deduplication keys. Search API availability does not mean full text is reusable.

Europe PMC API documentation ↗

Publication policy

No unsupported aggregate claims.

Prevalence, incidence, evidence strength, and country totals remain unpublished until the metric, source, inclusion criteria, reference date, and limitations can be shown together.

Contribute responsibly

Three ways to help the evidence grow.

Suggest an Africa-centred publication, join the scientific review process, or improve the open ingestion and interface code. Do not submit patient-identifiable data.