NeuroMetBench reproducibility package
Description
NeuroMetBench is a patient-aware evidence-bounding methodology for evaluating transcript-derived metabolic representations across heterogeneous proteogenomic contexts. Version 2.1.2 adds revision-stage patient-influence and pathway-specificity stress tests, an internally preanalysis-frozen orthogonal measured-metabolomics sensitivity in the Wang/CPTAC glioblastoma cohort, and a recovered reaction-identity/artifact-integrity firewall for historical METAFlux outputs. The new evidence is deliberately asymmetric: LSCC pathway specificity is supported exploratorily, CNS specificity replication is not supported, and correspondence between the transcript-derived serine score and measured L-serine remains weak and uncertain. Historical solver optimality/KKT evidence is not reconstructed, and prior adverse results and claim ceilings remain preserved. Licensing is mixed by file: NeuroMetBench-authored material is MIT-licensed, while the minimal redistributed cBioPortal/Wang-CPTAC input subset remains under ODbL 1.0. The Zenodo record should declare both licenses.
Notes
This archive is the complete reproducibility release for NeuroMetBench v2.1.2. It contains the manuscript and Supplementary Information sources and PDFs, manuscript figures, executable scientific and provenance verifiers, reproducibility software, revision-stage robustness and sensitivity evidence, recovered historical numerical artifacts, and the immutable historical v2.0.7 release required for provenance.
The principal external endpoint is same-gene RNA-protein rank concordance and must not be interpreted as measured metabolic flux, causal validation, clinical validity, or universal metabolic-model validation. Version 2.1.2 preserves unfavorable and unsupported findings, including weak and uncertain correspondence with measured L-serine, unsupported CNS pathway-specificity replication, absence of historical solver optimality/KKT certificates, and the previously identified reaction-order blind spot.
The release uses closed-world SHA-256 verification and includes executable checks for scientific results, provenance, package integrity, software behavior, and figure regeneration. From the extracted release root, the outer archive can be checked with python reproducibility/verify_outer_release.py.
Licensing is file-specific. NeuroMetBench-authored material is released under the MIT License. The minimal redistributed cBioPortal/Wang-CPTAC input subset required for offline reproduction of the measured-metabolomics sensitivity remains governed by ODbL 1.0. See LICENSE_SCOPE.md and THIRD_PARTY_NOTICES.md inside the archive.
Files
NeuroMetBench_v2.1.2_reproducibility_release.zip
Additional details
Related works
- Is supplement to
- Software: https://github.com/build33Mercury/NeuroMetBench (URL)
Software
- Repository URL
- https://github.com/build33Mercury/NeuroMetBench