Open noahharrison64 opened 2 years ago
Hi,
To me, I imagine the results were be reasonably similar, or am I missing something crucial?
Depending on how flexible your protein is, the results may not vary too much. The main reason you want to "fix" the position of the protein is to remove global rotation and translation of the protein, so that the analysis of the water on the grid occurs in the reference frame of the protein. As you have discovered, this can be done after the fact via RMS-fitting. There used to be a problem with RMS-fitting and imaging in that imaging could not be properly done after RMS-fitting, but this was fixed as of version 5.0. However, the PME version of GIST is still not compatible with coordinates that have been RMS-fit (issues with the reciprocal calc.), so if you want to use PME-GIST you will want to restrain your protein.
I am currently testing code that rotates the grid instead of the coordinates, which will make it so that simulations with restraints are no longer necessary, but for now if you want to use PME-GIST you'll want a restrained simulation.
@fwaibl and/or @tkurtzman may also have comments here. Hope this helps.
The stat. mech. equations used for the entropy are only exact for a fixed protein (fixed external field). If you allow the protein to move, it will shmear out the distributions and give an underestimate of the entropy of the ordering. That being said, not allowing the protein to move has its own set of problems in the model system. In some circumstances, the water structure and energetics is very sensitive to small fluctuations in the surface. In my group, we tend to run with several different degrees of restraints in the simulations... All heavy atoms restrained, backbone only, and three sites restrained. For the three sites restrained we find 3 sites far from the region of interest (generally the binding site) and keep them harmonically restrained. We tend to choose these 3 sites from atoms that have low b-factors. This allows a good deal of natural flexibility in the protein while effectively preventing the protein as a whole from rotating and translating. Presumably it should be okay to allow the protein full flexibility (no restraints) and align the structures. This, of course, introduces some variability from how the structural alignments are executed. When we've done this, we've tended to focus on the region of interest and align using residues in that region (to get the best alignment for where we are interested in).
Dan, I wasn't aware that PME-GIST doesn't work with the RMS fitting... though maybe you told me and I forgot... Is this anything that we could deal with in my group?
Thanks for the info both of you. I have had some success with unrestrained but RMS fitted trajectories using AmberTools22. I don't know whether the PME version is included in this package. I'm also running backbone restrained simulations, so will see if I get wildly different results.
I suppose an interesting aspect arises when considering the energetics of water displacement by ligand binding. If you were to use a RMS fitted trajectory, then large sidechain fluctuations will likely mean that superposition of the native crystal ligand no longer represents an accurate binding mode. Therefore, you'd probably have to re-dock the ligand. However, if you fully restraint the protein and then simulate, then the native ligand should fit perfectly into the structure.
Hi,
I've been running MD simulations using OpenMM and want to carry out water analysis using GIST. While I haven't restrained the protein using strong position restraints, as suggested by the tutorial, I did fit the protein in the output trajectory to the positions in the initial frame. To me, I imagine the results were be reasonably similar, or am I missing something crucial? I also noticed this discussion and was wondering if there is any update on GIST tutorials?
Thanks, Noah