Similar to the Rosetta simulations done in Kidd, Baker,Thomas "Computation of Conformational Coupling in Allosteric Proteins", PLoS Computational Biology , 2009, 5(8), I would like to generate structure ensembles for allosteric regulated proteins using PyRosetta. Starting from crystal or homology structure with/without a bound allosteric effector molecule I want to generate new conformations to analyze the effect of different compounds. In the article the "rebuild and refinement" sampling methodology is mentioned and it is referred to Rohl et al, Proteins 2004, 55, 656, and Qian et al, Nature, 2007, 450, 259.
Unfortunately, I do not know how to set up or where to find this script, respectively, using the described 3 steps (generating structural diversity, optimiying side chain positions, minimizing all atoms). First, I started with the ClassicRelax and FastRelax functions, but these scripts are computationally too expensive (6h and 30min, respectively, for one conformation of a ~500 residue protein) to calculate hundreds of conformations.
David, Jeff and Sergey pointed me to the LoopRelaxMover (accessible by 'import rosetta.protocols.loops' and then creating it by using 'm = rosetta.protocols.loops.LoopRelaxMover()'; afterwards, it can be applied using ie 'm.apply(pose)'). Nevertheless, it is not clear to me how to apply this mover in combination with Monte Carlo/minimzation to get the described "rebuild and refinement" script. Furthermore, where can I find a documentation and possible parameter settings for the available rosetta protocols (eg. ClassicRelax, FastRelax, LoopRelaxMover,...)? Using the normal binary distribution (PyRosettaUbuntu-r40224.linux.32Bit/PyRosetta.ScientificLinux-r40251.linux.64Bit) the source code for these rosetta protocols is not available.
Thank you very much for your help!
Stefan
Are you still interested in this?
As far as FastRelax being slow: All of the relax modes are slow on proteins of greater than 100 residues or so. PyRosetta won't make that any faster (it will probably make it slower).
As far as source code: You can get the C++ source under the academic license; PyRosetta is mostly a python interface to the same C++ source.
As far as documentation: We know it is bad, but mostly what you are looking for is not there. We are (eternally) working on it.
As far as not knowing how to write the script: I'm not sure quite what you need. It seems that you are proposing a very extensive new version of the relax code, or maybe just a python interface to the relax code?