Redocking vs Cross-Docking: Test a Protocol Across Receptor Conformations

Before committing a library to a docking workflow, a researcher needs to know what the controls actually tested. One attractive redocking pose can hide sensitivity to the search seed, preparation or receptor state. This worked example retains four runs, including unsuccessful ones, to demonstrate a reviewable control rather than claim a validated production protocol.

The example uses mouse c-Abl chain A from PDB 1IEP, which contains imatinib, and PDB 1M52, which contains PD173955. The accompanying structural study describes different activation-loop conformations. Docking imatinib into the prepared 1IEP receptor is the redocking control; docking the same prepared ligand into 1M52 is the noncognate control.

The original 1IEP imatinib coordinates provide the reference pose. There is no imatinib experimental pose in 1M52 in this comparison. Distances to the transferred 1IEP reference therefore describe similarity to that cognate pose in an aligned frame, not a measured error against an experimentally observed imatinib–1M52 complex.

For this deliberately bounded control, each receptor retained chain A protein atoms; waters, bound ligands and other hetero records were removed. Alternate locations were limited to blank or A. Meeko 0.7.1 prepared both receptors without allowing silent omission of unresolved residues. The same prepared imatinib ligand was used in all four runs.

Before preparation, 1M52 was superposed onto 1IEP using 267 Cα pairs with matching residue numbers and residue identities. The fit RMSD was 4.973 Å. That is a large global mismatch: the comparison is sensitive to the alignment definition and should not be treated as a clean isolation of receptor-state effects. A production assessment should inspect the local binding-site fit and justify its alignment atom set before interpreting cross-docking RMSD.

We retain this limitation because it changes the conclusion. Receptor geometry, the alignment, rigid-receptor treatment and preparation can all contribute to a poor cross-docking result. Removing cofactors or waters is a stated choice in this example, not a general recommendation for every target.

The official Vina basic-docking tutorial notes that imatinib can be difficult at default search effort and discusses increasing exhaustiveness. This example deliberately records the results at exhaustiveness 8. Two seeds are a diagnostic check, not a sufficient sampling study or a basis for estimating a reliable success rate.

The top-ranked redocking pose for seed 43 closely reproduces the reference under this calculation. Seed 42 does not: even its best returned pose remains 6.605 Å away. Reporting only seed 43 would hide a search-stability problem already present before changing the receptor.

Neither cross-docking run recovered a pose close to the transferred reference. That result is specific to these preparations, alignment, box, scoring model and search settings. It does not show that imatinib cannot bind a biological state, and the numerical score differences are not an experimental affinity comparison between receptor conformations.

The output poses were converted through Meeko and compared using chemical identity checks, symmetry-aware atom mapping and fixed-frame heavy-atom RMSD. The ligand was not independently fitted onto the reference. Fitting the ligand itself can conceal that it occupies the wrong place in the receptor.

Use the atom-mapping and redocking RMSD guide for the calculation details. In cross-docking, the receptor alignment adds another layer: a precise ligand RMSD is still conditional on the coordinate frame you selected. Always retain the alignment transform and the atoms used to compute it.

First make the cognate control reproducible across a justified search budget and additional prespecified seeds. Then examine the local receptor alignment, ligand state, pocket geometry and preparation choices. Change one defined aspect at a time, preserve the original results, and state which question the revised experiment addresses.

For a broader evaluation, use several suitable ligand–receptor pairs and keep protocol development separate from a held-out assessment. If the intended task is ranking a library, add a ranking evaluation with justified labels and controls; pose recovery and enrichment answer different questions. The docking validation guide connects those checks.

MolNexus provides receptor and ligand preparation, docking configuration and execution, pose review, local history and result export in a Windows desktop workspace. Organize the receptor-specific jobs and retain the exported settings and results. Receptor alignment, cross-docking study design and the RMSD calculations described here remain explicit analysis steps; this is not an automatic MolNexus cross-docking benchmark.

For evaluating the desktop workflow, MolNexus currently offers 15 free ligand dockings. The allowance is counted per ligand submitted, not per multi-ligand job. Review the product page for the current Windows requirements, license and purchase terms.

Explore MolNexus