PyRosetta Integration

ParmEd provides support for the PyRosetta API. The ability to create a de novo molecular system and then seamlessly parameterize and simulate it, all within the Python environment, is powerful and fits right within ParmEd’s goal of easing the molecular modeling process.

This page is not meant as a tutorial for the PyRosetta library. Instead, you should visit the PyRosetta website for that. However, this page will provide a brief description of PyRosetta, and how ParmEd can enhance its utility for molecular modeling. It will also present a few examples using PyRosetta with the tools provided by ParmEd.

What is PyRosetta

Rosetta is a popular molecular modeling suite that allows scientists to model biomolecular systems, and has had a lot of success in predicting experimental structures. PyRosetta is a Python-based interface to the Rosetta, and it allows users to easily model and manipulate biomolecules, enabling custom algorithms for protein folding using Rosetta sampling and scoring functions. Its features include:

  • Generating a nucleic or amino acid structure from a text string.
  • Mutating residues in an existing structure.
  • Manipulating residue positions and torsions in an algorithmic fashion.
  • Monte Carlo sampling using the Rosetta force field.

How does ParmEd enhance PyRosetta?

load_rosetta(pose) Load a Pose object and return a populated Structure

ParmEd provides a simple function, load_rosetta(), to load a PyRosetta Pose object into ParmEd’s Structure class. This Structure can then be used to easily parameterize a molecular dynamics simulation.

One can imagine that this functionality would be useful to easily seed simulations for ab initio structure predictions or studies of a mutant protein using wildtype structures.


For examples on how to run an simulation starting from a PyRosetta Pose please refer to the following:


One minor complication with using PyRosetta to prepare a structure for simulating with OpenMM is that proline residues have virtual sites that prevent it from being recognized by OpenMM. If you get an error that looks like the following:

Traceback (most recent call last):
  File "", line 11, in <module>
    positiveIon='Na+', negativeIon='Cl-', ionicStrength=0.1*molar)
  File ".../python2.7/site-packages/simtk/openmm/app/", line 372, in addSolvent
    system = forcefield.createSystem(self.topology)
  File ".../python2.7/site-packages/simtk/openmm/app/", line 785, in createSystem
    raise ValueError('No template found for residue %d (%s).  %s' % (res.index+1,, _findMatchErrors(self, res)))
ValueError: No template found for residue 2 (PRO).  This might mean your input topology is missing some atoms or bonds, or possibly that you are using the wrong force field.

then you know that you have to strip all NV atoms (which are proline virtual sites) from the Structure. You can do this by calling .strip('@NV') on your Structure prior to building the OpenMM System object.