DockPrep Audit

DockPrep Audit is an open-source, local-first quality-control tool for receptor structures used in molecular docking. It does not alter a structure or claim that it is biologically correct. Instead, it makes preparation choices visible before a docking run.

What v0.2 audits

Whole-structure findings (any occurrence anywhere in the receptor):

• Alternate atom locations (ALTLOC_PRESENT): a receptor requires an explicit conformer policy.
• Water molecules, non-standard residues, and metal atoms (WATERS_PRESENT, METAL_PRESENT): potential protocol-dependent decisions.
• Missing occupancy and missing element annotations.
• Duplicate atom records within a residue.

Site-local findings (scoped to the declared ligand's binding site, added in v0.2 — require a ligand argument):

• SITE_BRIDGING_WATER_PRESENT: a water within hydrogen-bond range of both the ligand and the receptor.
• SITE_ALTLOC_PRESENT: a receptor alternate-location conformer near the binding site (excludes disordered waters and the ligand's own atoms).
• SITE_METAL_PRESENT: a metal ion near the binding site.

The output is a machine-readable JSON report and a portable HTML report. The structure stays local; no data are uploaded.

Companion study

manuscript/draft.md and benchmark/ contain a 60-case redocking benchmark testing whether these findings actually predict preparation-policy sensitivity (water retention, altLoc resolution, metal retention), including a 76-case independent validation cohort and a formal predictive evaluation. PROJECT-ROADMAP.md section 19 documents the full methodology and results (items F1-F9).

Quick start

py -m pip install -e .
dockprep-audit audit path\to\receptor.pdb --output audit-results

Open audit-results/report.html in a browser. Exit status is 0 when no high-severity finding is present and 2 otherwise.

Scientific scope

This is a preparation-audit tool, not a docking engine and not an automatic repair program. Its intended contribution is traceability: before any repair, deletion, protonation, or conformer selection, the user receives an explicit report of the decisions that can affect a docking workflow.

Status

1. Validate rules against curated DUD-E/PDB receptor examples. Done (102-target audit, 60-case redocking benchmark).
2. Quantify rule prevalence and inter-protocol disagreement. Done (manuscript/draft.md Sections 3.5-3.10).
3. Add a graphical interface and exportable provenance manifest. Not started.
4. Archive a release in Zenodo and submit the manuscript. Manuscript drafted, pending final author/DOI details (manuscript/draft.md, [DATA:] markers); Zenodo archive and GitHub repo not yet created.

License

MIT. See LICENSE.