PDB Retraction Audit: Status Alignment, Retraction Signalling, and Reproducibility Dataset

A reproducible dataset and research package for an audit of how Protein Data Bank (PDB) records respond when their associated primary research publications are later retracted. The frozen cohort contains 43 retracted primary publications linked to 81 unique PDB entries. Each publication and entry was evaluated using primary retraction evidence to determine whether the deposited structural result was directly implicated, not identified as affected, or unclear. PDB entry status and user-facing retraction signalling were then assessed at both publication and entry level. Key findings: All 37 entries directly implicated by retraction evidence were obsolete. All 23 entries that remained released were linked to cases where the structural result was not identified as affected. None of those 23 released entries displayed explicit PDB-controlled retraction context. Obsolete-entry retraction signalling was present for 46 of 58 obsolete entries. The deposit includes frozen raw and processed data, adjudication records, exclusion logs, analysis outputs, figures and tables, supplementary data, protocol documentation, reproducibility code, environment specifications, manifests, and checksums.