Post-outcome reproducibility package: KORE-Map herbal-formula algebra (Axis A/B)

This is the post-outcome reproducibility package for the outcome-blind analysis of KORE-Map herbal-formula transcriptomes that was prospectively preregistered at DOI 10.5281/zenodo.21360321. The KORE formula-treatment outcomes were opened only after that preregistration had been publicly published; this record releases the resulting analysis, code, tests, and provenance. Prespecified decisions: Axis A STOP and Axis B STOP. The simple additive model, in which yield- and dose-adjusted constituent-herb responses are summed to predict held-out whole-formula responses, was not supported. The remaining large non-additive transcriptomic residual also did not pass held-context transport evaluation and Holm-adjusted confirmatory inference. Both prespecified axes therefore stopped under the frozen thresholds q_A = 0.20, delta_A = 0.05, m_B = 1.30, and t_B = 0.45. This record does not demonstrate clinical efficacy, safety, therapeutic synergy, or any human mechanism, and makes no claim of patient benefit. This record contains a reconstructed download copy generated from the verified public preregistration inputs after the original chat-delivered archive became unavailable. The analysis was re-executed under the frozen dependency versions. The prespecified conclusions remained Axis A STOP and Axis B STOP. Provenance: the primary results file results/AB_ANALYSIS_RESULTS.json has SHA-256 2ab63a7268ca0b5fc97fb0cd69d513b4cedaccb9d642e9be70d3e3a36f011019 and a size of 5,826,840 bytes. The archive additionally carries an internal SHA256SUMS.txt ledger covering every packaged file. Raw expression matrices and MSigDB gene-set membership data are not included or redistributed in this package. Licensing: this record is released under CC BY 4.0. The code contained in this package is licensed under the MIT License.