NeuroMetBench v2.0.7: Reproducibility and Provenance Package

Version 2.0.7 is a corrective synchronization and reproducibility release.

The validated scientific conclusions, cohorts, endpoints, thresholds, statistical results, canonical intervals, adverse findings, and interpretation ceilings are unchanged from the frozen scientific state.

This version corrects publication and reproducibility metadata identified during the final pre-submission audit, including:

synchronization of GitHub repository version metadata, citation metadata, Zenodo metadata, changelog, release notes, licensing records, VERSION, and checksum records;
correction of the exact supported-domain moderate-effect simulation value to 0.48705882352941177 = 6210/12750;
correction of reviewer-facing spatial terminology to 8/12 genes covered, corresponding to coverage 0.667, below the frozen 0.80 admission threshold;
narrowing of CNS wording from broad “generalization” language to a development-disjoint external-case association/reproduction, without implying independence of institution, platform, acquisition ecosystem, or upstream processing;
strengthened claim-evidence traceability and cross-package identity verification;
deterministic double-build verification;
repeated fresh-extraction stranger reproduction;
interruption-safety testing;
independent statistical recomputation;
public-package hygiene preserving historical identities and adverse statuses while excluding development-only workflow residue.

No adverse, null, unsupported, contradictory, or non-admitted result was removed or converted into positive evidence.

Important interpretation boundaries remain in force:

RNA-protein concordance is not direct measurement of metabolic flux;
the three-gene serine aggregate does not outperform the strongest constituent gene;
broad simulation calibration was not earned;
the spatial branch remains structurally non-admitted;
METAFlux numerical evidence is limited to the documented primal-feasibility compatibility assessment;
the CNS cohort is patient-disjoint but should not be described as institution-, platform-, or ecosystem-independent validation;
the study is a retrospective case study and does not establish prospective clinical utility or general validation of transcript-derived metabolic inference.

The reproducibility archive passed the complete 20/20 verification suite, deterministic A/B build identity, three independent fresh-extraction stranger runs, nested manifest verification, interruption-safety testing, and independent recomputation of the principal statistical authorities.

Raw third-party molecular matrices are not redistributed. Contemporary raw-to-final reproduction remains conditional on source availability, historical software environments and dependencies, and applicable external data-use and licensing terms.

Frozen reproducibility archive

Filename: NeuroMetBench_v2.0.7_reproducibility_release.zip

SHA-256: 955bfafae9bf2a3b85f06a0485f27760e23ab72c669ffedf061b736c90670bba

Size: 21,785,165 bytes

Historical releases remain immutable and are preserved as part of the project provenance record.