TM-align structure superposition doesn't reflect the underlying MSA

[apcamargo]

I've noticed that the induced superposition generated by TM-align does not perform well when a flexible linker is present between two domains. For example, in the superposition of two terminase structures shown below:

The endonuclease domain (top) aligns "too" well, causing the ATPase domain (bottom) to become misaligned. When three or more structures are superimposed, the endonuclease domains remain well aligned, while the ATPase domains are scattered. This misalignment does not reflect the underlying MSA, where the ATPase domains are properly aligned (which is what I expected from an "induced" superposition).

To address this on my end, I wrote a script that selects MSA columns without gaps and uses them as anchors to rotate the structures. The resulting superposition is shown below:

I don't think this is necessarily a bug, but I wanted to document it in case others come across this behavior.

[martin-steinegger]

Very good idea, we will look into it! Thank you so much.