Interface clustering problems in structures with widely variating NCS-related chains

[josemduarte]

In structure 1gxd, interfaces 1 and 2 should be clustered but they aren't since the rmsd must be above our 2 A threshold. PISA does not cluster them either.

The problem here is that the NCS-related chains (A,B) and (C,D) vary quite a lot in several loops, plus there's also some displacement between the 2 domains of (A,B).

A possible fix can be to measure rmsd only on contacting region.

See also related issues #1 and #2