Pose RMSD
Measures how far docked poses sit from a reference pose, in Ångström, symmetry-corrected and without superimposing. Use it to check whether a docking setup can reproduce a binding position that is already known — the standard way to decide whether to trust docking results on a target at all.
Pose RMSD
What it does
Takes a set of docked poses and one reference pose, and reports how far each pose is from the reference. One number per pose, in Ångström.
Almost everything else you can do with a docking result interprets it — which residues the pose contacts, how the score compares, whether it survives a simulation. All of that assumes the pose is roughly in the right place. This is the tool that checks the assumption.
Redocking: why this is the important one
The usual way to use it:
- Take an experimental structure with a ligand already bound.
- Extract that ligand (Ligand extraction) — this is the reference, the answer nature gave you.
- Remove it from the structure (Structure cleanup, Remove heterogens).
- Dock a fresh copy of the compound back in (Small-molecule docking).
- Measure the result against the reference — this tool.
If the top pose comes back within 2 Å, the box, the preparation and the engine reproduced a known answer, and you have a reason to believe results for compounds whose answer you don't know. If it does not, you have learned something considerably more valuable than a docking score: the setup is not ready.
2 Å is the field's convention, and it is reported in every result rather than left for you to remember.
Reading the output
| Column | Meaning |
|---|---|
| Pose | The pose's rank as docking returned it. Pose 1 is the top-scored one. |
| RMSD to reference (Å) | Distance from the reference pose. |
| Within 2.0 Å | Whether this pose counts as reproducing the reference. |
The summary separates two numbers that are easy to conflate:
- Top pose RMSD — how far pose 1, the one docking ranked best, is from the reference. This is the one that decides whether the redocking succeeded.
- Best RMSD — the closest pose in the set, whatever its rank.
When those differ, the tool says so, because the difference is itself a finding: if pose 4 is 0.6 Å and pose 1 is 5 Å, the search found the right position and the scoring function failed to rank it first. That is a scoring limitation, it is common, and it means you should treat that target's scores as a weak ranking signal rather than a measurement.
Two things it deliberately does differently
These matter because the alternatives look reasonable and give wrong answers that look fine.
It does not superimpose the poses first
There is a well-known way to compare two molecules that aligns them before measuring. That answers "is this the same shape?". Here both molecules are already in the receptor's frame, and the question is "is this in the same place?" — so aligning would destroy exactly what is being measured.
Concretely: a pose docked 5 Å away, into a different site entirely, measures 5.0 Å here and 0.0 Å if you align first. The aligning version reports perfect agreement for a completely wrong answer.
It corrects for symmetry
Many molecules have atoms that are chemically indistinguishable — the two oxygens of a sulfonamide, the two ends of a symmetric ring. Matching atoms by their position in the file will pair them up the wrong way round and invent error that isn't there.
Acetazolamide, with only its two sulfonamide oxygens swapped and the molecule otherwise untouched, measures 0.0 Å here and about 0.99 Å by naive atom-order matching — nearly half the 2 Å threshold, from a molecule that has not moved.
Limitations
- Heavy atoms only. Hydrogen positions depend on how each side was prepared, not on the pose, so including them would make the number depend on your protonation settings.
- Same compound on both sides. There is no atom-to-atom correspondence between different molecules, so the run fails rather than returning a number that means nothing.
- The reference needs 3D coordinates. A flat 2D SDF has no pose to compare against. If you compare docked poses to a 2D PubChem download by mistake you will get a large number and a warning, not a silent wrong answer.
- It measures position, not quality. A pose can be 0.5 Å from the reference and still be chemically implausible, and a large RMSD on a novel compound is not a failure at all — it is only meaningful when the reference is known to be correct for that compound.
Not the same as trajectory RMSD
Trajectory stability also reports an RMSD. That one is measured per frame across a molecular dynamics run, aligned on the protein, and answers "does the pose hold together under simulation?" This tool answers "was the pose right in the first place?" Different question, different stage.