Phase 5 Success Criteria

Checks for whether Siftome can move users from a gene or target to relevant public dataset triage without pretending to validate a biological mechanism or drug target.

Can users start from a gene or target symbol?

met

The `/genes` workflow accepts conservative gene symbols, stores the original query, and shows the normalized gene ID.

Can users audit synonym handling?

met

Only trusted exact synonym mappings are accepted, unknown phrases are rejected, and the page shows the matched term and source.

Can users get from a gene to dataset candidates?

met

Curated gene-to-dataset hints and dataset triage backlinks point users to ranked GEO dataset candidates or back to ranked results when no direct match exists.

Can users see evidence source boundaries?

met

The gene snapshot separates dataset metadata matches, curated normal tissue evidence, curated disease-context evidence, processed expression availability, and user-selected dataset evidence with source-specific confidence labels.

Can users see cautious biological context before selecting datasets?

met

Supported genes show normal tissue warnings, disease-context public evidence, and common trap warnings without changing dataset ranking or making validation claims.

Can users turn a gene snapshot into a triage project?

met

Gene snapshots can create a focused project from a normalized gene and user-provided biological context without automatically selecting datasets.

Can users export the gene snapshot evidence?

met

CSV and Markdown gene snapshot exports include dataset links, source separation, confidence labels, trap warnings, caveats, and source links.

Does the product avoid target-validation overclaims?

met

An executable wording guard rejects prohibited biological overclaims and the UI uses cautious language such as public expression evidence summary, dataset evidence available, and further analysis recommended.

Does Phase 5 include dataset-specific expression analysis?

partial

Phase 5 intentionally stops at public dataset triage and cautious source-backed context; processed expression summaries and user-selected dataset expression evidence are still marked unavailable.

Remaining gaps

Add real processed expression availability and dataset-specific expression evidence only after the source data and analysis handoff boundaries are implemented.

Keep gene snapshots out of target validation, mechanism proof, druggability, and clinical actionability.