Reliability-Aware Explainable QSAR and Orthogonal Virtual Screening for MMP-1 Inhibitor Prioritization: reproducibility archive

Reproducibility archive supporting the manuscript ‘Reliability-Aware Explainable QSAR and Orthogonal Virtual Screening for MMP-1 Inhibitor Prioritization’. Version 3.1 preserves the frozen scientific source files from v3.0: the curated 1,911-compound dataset, descriptor matrix, scaffold-grouped predictions, 26-node KNIME workflow, Python scripts, screening records, importable Pharmit queries, docking records, environments and checksums. Node 8 primary performance (pooled OOF R² = 0.589472) remains distinct from the external Windows reference (R² = 0.585106) used with 60 matched Y-randomization replicates. Linux refit and single-fold repeatability evidence retain their documented limitations. The PubChem comparison is descriptive, not independent external validation. This version adds audited PyMOL complex visualizations and editable sessions, hybrid orthographic PyMOL/matplotlib pharmacophore figures, Williams leverage diagnostics, scaffold-cluster versus row-bootstrap intervals, descriptor-collision audits, and synchronized manuscript and supplementary review snapshots. These additions required no new QSAR training, permutation or docking runs. Image integrity and provenance checks accompany the rendering materials; recorded rendering repeatability is environment-specific. Consult README.md and REPRODUCIBILITY_GUIDE.md for execution instructions and limitations. The enclosed editorial snapshots and release-status files record the preparation stage; their references to an unpublished v3.1 or to the preceding DOI are historical provenance, not the current public availability status of this record. Funded by Prince Sattam bin Abdulaziz University, project PSAU/2026/03/45334.