Structure alignment
Compares two structures and reports TM-score and RMSD, to judge whether one can stand in for the other.
Structure alignment
What it does
Superimposes two structures and reports how similar their folds are. The typical use: you have a wild-type complex that defines an interface well, and a mutant structure that reflects the surface you actually care about. Aligning them tells you whether it's defensible to use one to interpret the other.
Runs on the USalign implementation, so the numbers are comparable to published TM-scores rather than an in-house approximation.
What the numbers mean
TM-score runs from 0 to 1 and, unlike RMSD, is normalised by protein length — so it means the same thing for a 60-residue domain and a 600-residue protein.
| TM-score | Reading |
|---|---|
| ≥ 0.9 | Essentially the same fold. Effectively interchangeable structurally. |
| 0.5 – 0.9 | Same overall fold with real local differences. Usable as a template, but check the specific region you care about. |
| < 0.5 | Different folds. Treating one as a stand-in isn't justified by this comparison. |
Roughly, TM-score above 0.5 means the two proteins share a fold; below 0.17 is what you'd expect from two unrelated structures.
TM-score is asymmetric. It's normalised by the length of one structure, so comparing a small domain to a large protein gives two different numbers. RxCubed shows both and treats the one normalised by Structure A as the headline value. If they differ a lot, the two structures are very different sizes — worth knowing before you read too much into either number.
RMSD is the average distance between matched atoms in Ångströms, over the residues that aligned. Lower is more similar, but it's length-sensitive and dominated by the worst-fitting regions — two structures can have an excellent TM-score and a mediocre RMSD because of one flexible loop. Read them together.
Aligned length is how many residues actually matched. A great TM-score over a small fraction of the protein is a much weaker claim than the same score over most of it.
What it doesn't do
- It doesn't tell you which residues differ — only how similar the folds are overall. Use the contact mapper for residue-level detail.
- It doesn't compare more than two structures at a time, and it won't suggest which pairs are worth comparing.
- There's no visual overlay of the superposition in this version. You can open each structure individually in the 3D viewer.
- A high TM-score says nothing about whether the binding behaviour is the same. Structural similarity is necessary, not sufficient.