When Assay Overlap Is Not Independent Evidence: Data, Code, and Provenance Audit Outputs

This repository contains the data, analysis scripts, provenance-adjudication outputs, summary tables, figures, and reproducibility files supporting the study “When Assay Overlap Is Not Independent Evidence: A Provenance Audit.” The study evaluates whether numerical cross-assay overlap in public bioactivity databases represents genuinely independent experimental evidence. A provenance-aware framework was applied across ChEMBL 37, BindingDB 202609, and direct PubChem BioAssay data, while treating assay compatibility and provenance independence as separate dimensions. The deposited materials include source-specific screening and filtering outputs, frozen provenance and assay-compatibility adjudications, PubChem network-connectivity analyses, repeat-adjudication quality-control files, manuscript-supporting tables, analysis scripts, JSON summaries, CSV datasets, and SHA-256 reproducibility manifests. Key analyses include the assessment of apparent versus provenance-independent assay-pair connectivity, source-lineage relationships across documents and discovery programs, and the effect of provenance filtering on the apparent evidence network. The repository is intended to support transparent reproduction, verification, and reuse of the computational workflow and study outputs. No human participants or live-animal experiments were involved. All analyses were based on publicly available bioactivity records.