Implicit Membrane Model

[Joshua Adelman]

I was wondering if it was possible to access the Z coordinate in the CustomGBForce class (or if there are future plans to allow users to access it), in order to implement a implicit membrane model. I have not played around with this class, but it does not appear possible given the current API.

[John Chodera]

I'll second the request for this! The implicit membrane GB models from Michael Feig and Charlie Brooks might be good ones to look at for examples of what might be allowed by the CustomGBForce class.

See, for example, the references under "Implicit Models of Heterogeneous Dielectric Environments" at the bottom of this page:

http://feig.bch.msu.edu/main-research-methodology.html

[Peter Eastman]

What would be required to support the models you're interested in? How about if computed values and energy terms of type SingleParticle (but *not* particle pair ones) could depend on the (x, y, z) coordinates of the particle?

Peter

[Peter Eastman]

Ok, I've just checked in this feature. Give it a try, and please let me know how it works.

Peter

[Nicholas Schafer]

Hi Peter -

I'm interested in having a ParticlePair CustomGBForce that depends on the z-coordinates of the interacting particles. Is it possible to implement such a force by using SingleParticle computed values that depend on the z-values of the interacting pair of particles inside the energy expression of a ParticlePair force?

Best,
Nick

[Peter Eastman]

Yes, that ought to work. You can have a SingleParticle computed value that depends on z. A ParticlePair energy term can then depend on the values of that computed value for the two particles.

Peter

[Nicholas Schafer]

Perfect. Thanks again for your efforts.

Cheers,
Nick

[Wei Lu]

I got an error
(OpenCL Error : clEnqueueNDRangeKernel failed: kernel_arg 6 has not yet been initialized) using openCL
(CUDA_ERROR_INVALID_VALUE) using CUDA, no error using CPU.
when I tried to implement this implicit membrane model,

Code: Select all

contact = CustomGBForce()
gamma_ijm = np.zeros((2, 20, 20))
contact.addTabulatedFunction("gamma_ijm", Discrete3DFunction(2, 20, 20, gamma_ijm.T.flatten()))
contact.addComputedValue("rho", f"1", CustomGBForce.ParticlePair)
contact.addComputedValue("alpha", f"z", CustomGBForce.SingleParticle)
for i in range(pdb.topology.getNumAtoms()):
    contact.addParticle()

But if I comment out one line

Code: Select all

contact.addTabulatedFunction("gamma_ijm", Discrete3DFunction(2, 20, 20, gamma_ijm.T.flatten()))

then no error will occur.

simplified example is attached here, you can run it with

Code: Select all

 python3 mm_run.py 

customGBforce.zip

(47.18 KiB) Downloaded 32 times

Please help.
Thanks a lot.

Wei Lu

[Peter Eastman]

I just posted a pull request with the fix. Could you try it out and see whether it now works for you?

https://github.com/pandegroup/openmm/pull/2248

[Wei Lu]

It works! thanks a lot!