pdicert: outcome-independent certification and selection for reproducible delivery of biomedical prediction scores — software and reproduction package

Software and reproduction package accompanying the manuscript “pdicert: outcome-independent certification and selection for reproducible delivery of biomedical prediction scores.” The archive contains Python source code, tests, locked configurations, retained predictions, reference results, provenance manifests, and documentation for outcome-independent prediction-delivery certification, minimum-byte certified text selection, and outcome-aware table–summary verification. Its default replay starts from retained predictions and checks 21 unit tests, 1,728 delivery cases, 13,824 checker decisions, 1,680 precision exports, and 48 selected representations; it does not refit the models. Full refitting requires separate downloads of the public GEO expression datasets documented in README.md; source expression matrices are not redistributed. The original software and accompanying original documentation are licensed under the MIT License by Chung-I Huang and Iris Yang. Third-party dependencies, datasets, and other materials retain their own terms.