Molecular Docking with Waters, Metals, and Cofactors: What Should Stay in the Receptor?

Protein structures contain more than amino-acid coordinates. Experimental entries may include bulk solvent, ordered waters, crystallization agents, buffer ions, catalytic metals, structural ions, substrates, products, cofactors, prosthetic groups, and bound inhibitors. Some are incidental. Others define the chemistry of the pocket.

The official AutoDock Vina basic-docking workflow says that waters, ligands, cofactors, and ions deemed unnecessary may be removed during receptor preparation [1]. The decisive word is “deemed.” Cleanup is a scientific classification step, not a command to delete every heteroatom.

RCSB PDB treats ions, solvent molecules, cofactors, inhibitors, substrates, and other non-polymers as ligands in the archive. Its ligand-quality guidance distinguishes functional and non-functional components and exposes agreement with experimental density, geometry, clashes, and comparison across entries [2]. Use that evidence before deciding that a three-letter residue code is either essential chemistry or disposable clutter.

Start with the broader protein-structure selection workflow so that component decisions remain tied to the intended biological state rather than to one coordinate file.

A water molecule can bridge ligand and receptor interactions, occupy an unfavorable site that a ligand displaces, stabilize a pocket network, or simply appear because the experimental map supported its placement. Studies of structural water show that incorporating well-placed waters can improve interaction scoring or pose prediction in particular benchmarks [5,6]. Those results do not create a universal rule to preserve crystallographic water.

Water positions can vary with ligand, protein state, resolution, temperature, and model-building choices. WaterDock, for example, was developed to predict water positions around protein-ligand interfaces and its analysis reinforces that observed conservation must be evaluated rather than assumed [6]. Treat each retained water as part of the protocol and identify it by chain, residue, coordinates, evidence, and intended role.

AutoDock Vina documents a specialized hydrated-docking method in which water-like dummy atoms are attached to the ligand, evaluated with a modified AutoGrid water map, filtered after docking, and rescored [3,7]. This is not equivalent to leaving all receptor waters in a standard Vina job.

The official tutorial states that the method was calibrated and validated with the AutoDock4 force field, advises against using the protocol with the Vina or Vinardo force fields, and says the current implementation is not suitable for virtual screening without additional normalization because scores across diverse ligands are difficult to compare [3]. Follow the documented method boundary rather than mixing one protocol's preparation with another protocol's scoring.

A catalytic zinc ion is not interchangeable with a monovalent salt ion. Metal identity, oxidation state, coordination number, geometry, coordinating residues, ligand donor atoms, protonation, and nearby waters can determine the binding hypothesis. Generic nonbonded treatment may not reproduce directional coordination chemistry.

AutoDock4Zn was created as a specialized AutoDock4 force field for zinc-coordinating ligands. The original method was calibrated on 292 zinc complexes, and the current Vina tutorial requires a dedicated parameter file, zinc pseudo-atoms, preparation scripts, and AutoGrid maps [4,8]. The lesson is not that zinc docking is impossible. It is that the protocol must explicitly support the chemistry being claimed.

A heme, flavin, nucleotide, or other cofactor may create the binding surface, define electrostatics, establish the redox or catalytic state, or stabilize the receptor. Removing it can create a pocket that does not represent the intended biology. Retaining it without correct connectivity, charge, protonation, atom types, and state can create a different artifact.

Confirm the component in the wwPDB Chemical Component Dictionary, inspect its local experimental support, determine whether it is covalently attached, and identify the biologically relevant state. Then verify that the preparation and scoring stack can encode it. If it cannot, choose a validated specialist method or narrow the claim rather than hiding the mismatch.

The general receptor and ligand preparation guide provides the surrounding checklist for cleanup, protonation, prepared inputs, and review.

When evidence does not settle whether a component should remain, create a small, prespecified comparison: for example, dry receptor versus one retained bridging water, or standard preparation versus a validated metal-specific protocol. Keep receptor identity, ligand set, box, search settings, seed policy, and evaluation criteria controlled wherever the methods permit.

Compare more than the most favorable affinity. Inspect recovery of relevant cocrystal poses, coordination or interaction geometry, known-ligand behavior, false-positive patterns, failure rate, and sensitivity across seeds or receptor structures. Select the simplest variant that answers the biological question and passes the frozen validation gate.

Define that gate before screening by following the molecular docking workflow-validation framework.

MolNexus 0.1.1 exposes receptor cleanup choices before PDBQT generation and keeps preparation review beside the interaction box, AutoDock Vina 1.2.7 execution with Vina or Vinardo scoring, pose inspection, local SQLite history, and exports. This visibility is useful for standard protein-small-molecule docking studies in which researchers must decide and record what is excluded.

The current product specification does not present specialized hydrated docking, AutoDock4 scoring, AutoDock4Zn, automatic cofactor parameterization, or metal-specific force fields as MolNexus features. Projects requiring those protocols should use the validated specialist workflow that represents their chemistry. That boundary preserves the value of a focused standard workflow without implying unsupported scientific coverage.

MolNexus is coming soon; purchase and download are not open. The 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.

Receptor cleanup should simplify the system without deleting the chemistry that makes the target relevant. Classify every binding-site water, metal, ion, and cofactor; inspect its evidence; match it to the ligand mechanism; and confirm that the selected force field can represent it.

Use a standard dry-receptor protocol when it passes validation. Escalate to selected fixed waters, hydrated docking, metal-aware parameterization, cofactor treatment, or receptor variants only when the biological question requires them. Record the decision so another researcher can understand not just what the receptor contains, but why.