I have two questions about using Modeller to build a chimera protein
based on two known structures (as in #1 of the FAQ):
1. How does Modeller determine the initial relative orientations of the
two structures?
The initial structure is the target sequence threaded onto the
template(s) backbone, so the initial relative orientation will be
essentially that of your input PDB files.
2. Is there a way to extensively sample the relative orientations of
the two structures? Perhaps through randomization of dihedral angles,
loop modeling, or some other approach?
Since you have no domain-domain restraints, Modeller will not be able to
do a good job of optimizing the structure. So you should impose some
sort of orientation restraint (maybe an angle between the mass centers
of the two domains and a central point). Then you could perhaps vary
that angle and build models to sample the orientation, but there are
many ways to do this.