™A number must show its ingredients
Each saved score includes its evidence classes, components, weights, model version, uncertainty, and missing-data effects. A reader should be able to see why a score changed and which inputs were measured, published, modeled, or absent.
Novelty and evidence depth need separate displays. An unexplored idea may have little evidence without being disproven. Contradictory studies should be shown, not hidden inside a single average.
Context controls interpretation
A signal for one cancer subtype, cohort, assay, or endpoint cannot simply be reused for another. Scientific calibration requires testing against known examples, negative controls, incompatible cohorts, and later outcomes before its claims expand.
A qualified researcher should still review the evidence, record a decision, and be free to disagree with the ranking. A low or missing score cannot be translated into a clinical recommendation.
Current status
Authorized project members calculate and inspect technically tested scores only when exact governed inputs pass the eligibility gate. Scores retain their revisions and reviews. A public illustration, traceable evidence lead, or software test is not scientific validation, and an empty project does not receive a fabricated score.