simtk.org

Dear all,

I've been trying to simulate a system of polymer chains in vacuum using internal variables. The plan is to constraint all bond lengths and angles, and to vary only several torsional angles. I've been trying to use Molmodel with SImbody for this purpose, because SImbody has Internal variable Module, which uses internal coordinates and allows to use constraint, implemented. However, I have to define the volume of the system (and to use NVT ensemble), which is not possible in Molmodel. Could, someone, please answer me if it is possible to use internal coordinates in OpenMM? I read on your forum that OpenMM doesn't use internal coordinates directly, but yit could be done with Molmodel (I already defined the structure of polymer in internal coordinates in Molmodel), and then OpenMM can be used for force evaluations. But I am not sure how to do this. Could someone, please, answer me is it possible to use both Molmodel (for internal coordinates), and OpenMM to define canonical ensemble?

Thank you very much for help.

Best regards,
Ljiljana

Hi Ljiljana,

OpenMM works in Cartesian coordinates, not internal coordinates. Depending on exactly what you're doing, that might or might not be a problem. You can still impose constraints on your system to make all the bond lengths and angles rigid. Is that the only requirement, or is there any other reason you need to work in internal coordinates?

Also, what exactly do you mean when you say you need to "define the volume of the system"? That implies you're using periodic boundary conditions, which is unusual for a vacuum simulation?

Peter

Thank you very much for the answer. Yes, I wanted to use the internal coordinates to constraint bond lengths and angles, that was the only reason. However, the user's guide states that it is only possible to use constraints for angles which involve H atoms. Is there maybe any way to constraint all bond angles?
The plan was to use periodic boundary conditions for the system of polymer chains in vacuum and to treat it as canonical ensemble, if possible.
Thanks once again.

OpenMM has no problem constraining anything you want: just call addConstraint() on your System to specify them. But the ForceField class doesn't provide any built in options beyond HAngles. There are two reasons for that. First, constraining additional angles in a protein simulation doesn't usually allow you to increase the time step, since the oscillation frequencies of those other angles are similar to the ones for whole water molecules. Second, when constraining all angles you have to be careful not to create an overdetermined system. If you just blindly add one constraint for every bond and one for every angle, some of those constraints can end up being redundant.

What sort of polymer are you simulating? Depending on the details, this might or might not ever come up.

Peter