Problem with SymmDock in helical symmetry

Keywords:
SymmDock
Helical Symmetry

Dear RosettaCommons,

we’re trying to model a protein complex in helical symmetry with SymmDock in Rosetta 3.5.
There is one problem occuring always, which we coundn’t solve so far: In most of the resulting
models there are single residues displaced from the backbone of the protein. Often the whole structure
looks completely broken. Relating to this we made the observation that the master subunit is not affected
but the strength of the effect (number of displaced residues) increases with the number of jumps between the master
subunit and the affected subunits. Also the distance of a displaced residue to the protein backbone increases with
the number of jumps between the slave and the master subunit. The problem is that these disrupted structures are often
scored as “good” ones because the displaced residues don’t clash with others. In addition to this, effected residues seem
to be always located at the potential interfaces between subunits. We found that this problem occurs during the high resolution refinement step
in fa-mode. E.g. giving the flag “-prevent_repacking” was completely avoiding these disruptions, meaning probably those occuring during the packing or repacking stage?
And the effect is somehow related to the movement allowed by the dofs controlling the rise and the rotation around the helical axis.
We generated the symmetry definiton file based on a template helical structure (once based on a homolog complex,
once based on a self generated “loosely packed” helix).
Also we modeled a 7-subunit system of the complex or a 21-subunit complex (controlled by the -r flag for “make_symmdef_file.pl”), this problem was always occuring.
All dofs in the sdf were properly ramdomized, as far as we know. But we found that not allowing rotation around z (the helical axis)
is reducing this effect strongly even though not completely. But of course for sampling the search space in helical symmetry
this dof is very important and removing it can’t be the solution.
Feeding in AtomPair distance constraints and/or an edensity map works very well but was not altering our problem.
The question is now how can we overcome this problem – is there any easy solution for it?
Is it maybe a problem with the symdef file?

Thank you already in advance!