PDB-AI case study: atom-level electron-density loss in X-ray radiation-dose series (datasets and analysis package)

Datasets and analysis package for the case study in "Protein Data Bank Atomic Insights (PDB-AI): provenance-aware integration of structural biology resources" (Sellitto, Rolfe and Snell). The case study followed the same atoms through published X-ray radiation-dose series and found that the carboxylate oxygens of Asp and Glu lose electron density with dose faster than the amide oxygens of Asn and Gln in the same structures. The analysis was developed on 23 series (134 structures) and replicated under a pre-specified plan in 33 independent series. Every dataset was built with PDB-AI v1.0.0 (doi:10.5281/zenodo.23572775) with every engine enabled. Files: pdbai_known_dose_atoms_public.csv.gz: PDB-AI output for the 134 structures of the 23 original series (682,311 atom rows; the columns empty in every row removed). original_134_build_records.zip: run summary with data dictionary, run manifest, entry list, and export, QC and coverage reports for that build. replication_pdbai_atom_site_dataset.csv.gz: PDB-AI output for the 351 structures of the replication series (1,090,479 atom rows). replication_atoms_analysis.csv.gz: the columns the replication analysis reads, without Asp/Glu residues refined at partial occupancy. replication_351_build_records.zip: run records, entry list and doses for that build. pdb-ai_case_study_v1.0.0.zip: the analysis package (a copy of PDB-AI v1.0.0, inputs, commands, environments, run records, scripts and outputs) that reproduces every result, figure and supplementary table of the case study. The published doses of the 23 original series are from Shelley and Garman (doi:10.5281/zenodo.5566557). The structural data derive from the PDB, PDB-REDO, wwPDB validation, UniProt and the AlphaFold Database, under their own terms.