Dissipative Particle Dynamics (DPD)

[Daniel Alejandro Matoz Fernandez]

Hi! I am new in this stuff. But I would like to know if it possible to apply a DPD (Dissipative particle dynamics) thermostat using openMM. On the other hand, I am trying to simulate a system compose by a molecule with to atoms. The think is there are intra molecular forces and extra molecular forces (between the molecules) Do you have any Idea how to do that using openMM.

PS: I wrote my own code, but I would like to improve the code using openMM as a lybrary in c/c++.

Best Regard

Daniel

[Lee-Ping Wang]

Hi Daniel,

I think DPD in OpenMM would be very cool, but I'm not sure how straightforward it is to implement.

I am not an expert in DPD, but I believe the main conceptual difference from MD or Langevin dynamics is that the random forces are acting on pairs of particles (in equal and opposite directions) rather than individual particles. Is this correct?

My initial impression is that this may be implemented as an Integrator class in OpenMM in the C++ layer, because the CustomIntegrator class only supports per-degree-of-freedom variables. DPD on the other hand appears to require per-pair variables.

Daniel, Peter: Let me know if this is accurate.

Thanks,

- Lee-Ping

[Daniel Alejandro Matoz Fernandez]

Thank you for your swift response. Indeed, you have a damping force + a random force. This ensure you have momentum conservation on the system. The damping force and the random force are usually linked via the dissipation fluctuation theorem.

In order to have the correct integrator you need to modify a litebit the velocity verlet integrator (Velocity-Verlet based integration scheme DPD–VV). See: P Nikunen,et. al. Computer Physics Communications, Volume 153, 3, 2003, 407-423 (http://lib.tkk.fi/Diss/2006/isbn9512280590/article4.pdf).

The other thing is I want to use the library for a simulation of living cell composed by to particles. So, there an intracellular potential and extracellular potential. It is possible to do that using the bonded force plus a force field defined by me?

I have my own code running, but i want to acelerate with OPENMM.

Best Regard and Thanks in advance

Daniel

[Lee-Ping Wang]

Hi Daniel,

Thanks for letting me know. It does sound like writing an Integrator class is the way to go, and it doesn't seem difficult although I don't know for sure. Peter is the main developer who wrote most of the code, and he can provide some helpful input.

As for the intracellular and extracellular potentials, I think it is straightforward to define external one-body potential functions. However, if the cell has a movable boundary, it might be more challenging. Let me know what you have in mind.

- Lee-Ping

[Daniel Alejandro Matoz Fernandez]

Dear Lee-Ping. My idea is to simple simulate a cell using two point particles in order to have cell division. That means, should be exist a intracellular potential that push away the two point particles within the cell until some critical distance. I been thinking that is possible to do that using a bonded potential like a molecule nad then analyze the distance of the particles in the molecule and decide if there are cell division or not. For me was easy to implement in my own code but I want implement in openMM for performance.

Than you

Daniel

[Lee-Ping Wang]

Hi Daniel,

You might be able to do this with a CustomNonbondedForce. This allows you to implement any functional form you want between any two particles without specifying that a bond needs to exist.

- Lee-Ping

[Peter Eastman]

The forces should be easy to do with custom forces of various types. The integrator will be more difficult. CustomIntegrator is designed for algorithms that integrate each particle independently, not ones that couple pairs of particles together. So I think that will probably require a plugin. At least it should be a pretty simple one as plugins go.

If you're interested in doing that, you should check out the following resources:

1. The OpenMM developer guide (http://docs.openmm.org/6.1.0/developerguide/index.html, also found in the docs directory if you download the prebuilt binaries).

2. The example plugin (https://github.com/peastman/openmmexampleplugin).

3. The existing integrators built into OpenMM to see how they're implemented.

And of course, I'll be glad to answer questions about anything that isn't clear.

Peter