Reply To: De novo loop modeling using pyrosetta

[Anonymous]

The best way to handle loops of different length is with different starting structures. Trying to handle multiple length loops in one Rosetta run is challenging; the residue numbering changes as you add residues and things get messed up.

The fastest way to make the starting structures is to create PDBs with the appropriate numbers of residues in the loops (either by deleting residues or copying the coordinates of one residue repeatedly), then run the loop modeling code with the build_initial option. (I don’t know how to access that by pyrosetta but it’s in the regular loop modeling executeable’s documentation someplace.)

Once you have those, it sounds like you want pretty much standard loop modeling with occasional design steps. You can pick apart the existing loop modeling protocols (or perhaps call them directly from pyrosetta), then separately integrate design steps. Design is easy, just run the PackRotamersMover with a PackerTask/TaskFactory that allows for design at the loop positions.

If you don’t already know python, it may be easier to just do this as repeated applications of the Rosetta loop modeling and fixbb executeables. You can iterate like so:
loop modeling (to create starting structures of different lengths)
fixbb (sequence diversity)
loop modeling (loop diversity)
fixbb (sequence diversity)
…repeat

It’s not nearly as fast as just coding design directly into the loop modeling steps would be, but it’s faster than spending several months learning to code in python and/or C++.