Can Your Molecular Docking Study Be Re-Executed? An AutoDock Vina Reporting Checklist

Docking results are unusually easy to summarize and unusually hard to reconstruct. The final table may contain only ligand names and affinities, while the calculation depended on receptor state, ligand chemistry, preparation software, box geometry, engine version, scoring function, search effort, and stochastic choices. Losing any one of those links can turn a precise-looking number into an uncheckable claim.

A July 2026 preprint audited 236 open-access SARS-CoV-2 main protease docking papers. Under its stated essential-field rule, 8.1% reached the level required for direct re-execution, and the search-box center appeared in 33.9% of the papers [2]. This is a recent preprint focused on one literature corpus, not a universal estimate of all docking research. Its practical warning is still useful: important execution details routinely disappear between the workstation and the paper.

These layers reinforce one another but remain distinct. A perfectly archived bad protocol is re-executable. A biologically plausible conclusion with missing files may be impossible to audit. The reporting target is therefore not a ceremonial checklist: it is a minimum executable record attached to a separately validated scientific design [1,2].

Record the AutoDock Vina version and how it was obtained. Also record the preparation tools and versions because a receptor or ligand can change before Vina sees it. Include the operating system, command-line binary or Python binding, relevant dependency versions, and any custom scripts or modified weights.

If the preparation record is incomplete, use the companion protein and ligand preparation checklist to recover the missing decisions before treating the protocol as executable.

The Vina Python API exposes the scoring function, CPU count, seed, receptor, ligands, box center, box size, spacing, and related configuration through its information methods [4]. Capture that machine-readable configuration when possible instead of reconstructing it later from memory.

A receptor PDB ID is provenance, not a docking input. Record the entry and assembly, then retain the exact downloaded structure and the final PDBQT. Document removed chains, ligands, waters, ions, cofactors, alternate locations, missing atoms or residues, protonation choices, charges, flexible residues, and any local repairs. The same principle applies to every ligand: retain its stable identifier or source representation, stereochemistry, protonation and tautomer choice, 3D-generation method, charges, rotatable-bond treatment, and final PDBQT [1,3].

For the search-space record, the grid-box and exhaustiveness guide explains why center, dimensions, and search effort must remain explicit study inputs.

Use cryptographic checksums for retained files. A checksum does not validate chemistry, but it establishes whether two people are holding the same bytes.

“The box covered the active site” cannot be executed. Report the center coordinates, dimensions, and units for every receptor. If the box came from a cocrystal ligand, residue selection, cavity detector, or manual placement, record that rule as well. The official Vina basic-docking example uses an explicit configuration with center and size values, and its output repeats the grid geometry [3].

For an ensemble, do not assume one coordinate triplet describes every aligned receptor. State the alignment reference and whether the box was transformed, recalculated, or reviewed independently.

A single fixed seed is useful for forensic repetition, but it does not measure robustness. If a conclusion depends on stochastic search, predefine a multi-seed experiment and report the seed list, aggregation rule, and variability. Do not run seeds until one produces the desired pose and report only that run.

Retain the stdout or log, output PDBQT, all requested poses, warnings, return code, elapsed status, and a manifest linking each result to its receptor, ligand, and configuration. Keep failures as first-class outcomes. Removing rejected or crashed ligands from the denominator silently changes the study population.

Then record every transformation between raw output and the published table: pose extraction, deduplication, clustering, rescoring, interaction filters, manual inspection, compound selection, and exclusions. FAIR principles emphasize rich metadata, provenance, and reusable data rather than files that are merely downloadable [5].

Move the package to a clean user account, workstation, virtual machine, or colleague who did not build it. Ask that person to follow only the README. Record whether installation succeeds, checksums match, the expected number of jobs runs, and the prespecified output checks pass.

Define tolerances before testing. Byte-identical output may be unrealistic across platforms or thread schedules, while the same top pose within a declared RMSD tolerance or the same rank band may be the scientifically relevant criterion. Explain which parts must match exactly and which may vary.

This handoff complements the broader workflow-validation process: validation asks whether the protocol is fit for the question, while clean-room re-execution asks whether another person can actually recover and run it.

MolNexus 0.1.1 is BioChemIntelli's local Windows desktop workspace for receptor and ligand preparation review, interaction-box definition, AutoDock Vina 1.2.7 execution with Vina or Vinardo scoring, pose inspection, local SQLite job history, and result export. Keeping those stages visible in one project reduces the number of context handoffs a researcher must reconstruct.

That integration does not make a study reproducible by itself. Researchers still need stable source provenance, a validated protocol, an external archival policy, declared tolerances, and a clean-room re-execution test. MolNexus is coming soon; purchase and download are not open. Its current commercial profile is a US$499 one-time license for one Windows PC at a time, perpetual use of the purchased version, and 12 months of updates.

Design the research record before the first production run. Give every source molecule a stable identity, preserve every prepared input, export the complete Vina configuration, retain raw successes and failures, and connect each published claim to the exact outputs and analysis that support it. Then give the package to someone who did not create it.

If that person cannot launch the intended calculation without asking what you meant, the study is not ready to be called re-executable. Fixing that gap before screening is cheaper than reconstructing it after publication.