Does anyone know if it is possible to perform a 3d alignment between
two proteins while only moving the selected atoms? (i.e. leaving the
unselected atoms for the alignment unmoved?)
Modeller's 3D alignment (superpose) does a rigid rotation and
translation of the entire protein. It doesn't really make sense to not
move the unselected part of the protein because that would distort the
structure. But you could certainly do the superposition and then combine
the moved selected atoms with the unmoved unselected atoms from the
original PDB file, either manually or with some simple scripting.