Code, frozen outputs, prespecification documents and reproducibility metadata for the manuscript:
Structural calibration and evidence boundary analysis reveal convergent molecular space with residual syndrome differentiation in treatment knowledge derived inference
This package is designed to be accurate about what is and is not reproducible from the supplied files. It does not relabel scaffolds as completed endpoint code and it does not silently replace frozen inputs.
| Manuscript layer | Supplied status |
|---|---|
| Treatment / target projection | Frozen code + inputs + outputs; clean-rerun key-table equivalence checked |
| Exact herb-target topology null | Frozen code + prespecification + outputs present |
| GO/KEGG functional calibration | Executed code + outputs present; third-party annotation snapshots not redistributed here |
| CAP disease correspondence | Frozen reconstruction/matched-null code + derived cohort inputs + outputs present |
| Psoriasis DCSMI validation | Code-frozen endpoint runner + outputs present; executed primary disease background = GSE147339 Tier-B |
| URTI DCSMI validation | Frozen outputs + exact replay/audit reconstruction + full functional reimplementation are provided. The lost historical generator and exact GSE63990/patient-level raw inputs remain unavailable, so historical null/bootstrap byte-for-byte regeneration is not claimed |
| Cross-layer contraction | Executed code/output present; portable adapter explicitly restores the functional-eligibility dependency |
- Master structural/null seed:
20260922where specified by the frozen analyses. - Phase 12A: 10,000 Curveball null graphs, 4 chains, exact row/column degree preservation.
- Functional calibration: 10,000 within-stratum structural permutations; 374-target universe; 19 matched strata.
- Psoriasis DCSMI: 10,000 structural permutations; 2,000 patient bootstrap resamples.
- STRING RWR secondary analysis: primary restart
r=0.5; sensitivity0.3/0.7;max_iter=10000; L1 tolerance1e-12. - URTI P3 is
NOT_IDENTIFIABLE / NOT_TESTEDunder a frozen pre-result amendment.
Exact frozen/executed source bytes are preserved under provenance/original_frozen_scripts/. scripts/*PORTABLE.py are convenience adaptations that change file locations (and, for cross-layer analysis, make an implicit functional-eligibility dependency explicit) without changing the statistical definitions. Every adaptation is listed with original/runnable SHA256 values in provenance/PORTABILITY_ADAPTATIONS.csv.
Author-derived frozen/derived inputs needed for audit are bundled under data/frozen_inputs/. Third-party database snapshots and licensed resources are not redistributed in this GitHub package; exact filenames and known hashes are recorded in data/manifests/EXPECTED_EXTERNAL_INPUTS.csv. The complete internal basic archive preserves the supplied snapshots separately.
The repository supports audit/reproduction of the current frozen respiratory/CAP/psoriasis workflow to the extent permitted by the supplied source files and external-resource availability. The exact historical URTI endpoint-generating source remains unavailable. v1.4 adds clearly labeled reconstructed code: an exact replay/audit script that reproduces all independently identifiable endpoint statistics from frozen assets, a full functional reimplementation for use if the missing raw inputs are recovered, and a forensic partial-D reconstruction. None is presented as the lost historical source.
See RUNNING_THE_REPOSITORY.md, KNOWN_LIMITATIONS.md, and MANUSCRIPT_ALIGNMENT_NOTES.md before public release.