Interaction energies of subsets

[Siddharth Srinivasan]

Hi all

Are there any plans in the near future to "officially" support energies between subsets of particles as a front line feature? The specific use case I am thinking about is a Monte-Carlo protein sidechain repacking scenario, where I compute the energies between a sidechain and the rest of the system, but don't need to compute the energies between two distant sidechains since they don't change. I am aware of a few ways to hack around this by using force groups, or zeroing out masses, or even simply computing the energy of the full system all the time. However as the system grows, and the complexity of the algorithm grows (by including backbone movements), computing full system energies at every stage is not viable.

I realize that this is primarily meant for MD simulations, but I am wondering if this is a requested feature by other groups as well.

[Peter Eastman]

How about using a CustomNonbondedForce with interaction groups?

Peter

[Siddharth Srinivasan]

Wouldn't I need to recreate the entire energy function manually in that case? I don't want to hand roll a LJ, electrostatics, GB myself if I can just use the energy functions built in...

[Peter Eastman]

Yes, although that's quite easy to do for Coulomb and Lennard-Jones. But GB isn't a pairwise additive force, so the interaction between two groups isn't well defined. That is, the interaction between two atoms is affected by the presence of all the other atoms around them. How did you intend to handle that?

Peter

[Siddharth Srinivasan]

Well the energy itself is pairwise decomposable, its the computation of the alpha radii that is not, right? That's why I was asking if there are any plans to support such a feature, where the appropriate alpha values would be updated, and then the energy could be computed on a subset. But I realize this is a very complicated implementation, so I would definitely not try and recreate this using a custom force. In the case of a simple nonbonded interaction, its much more feasible, but if you have GB and hydrophobic terms in there, it gets impossible.

[Peter Eastman]

Well the energy itself is pairwise decomposable, its the computation of the alpha radii that is not, right?

The Born radii are just an intermediate step in the computation of the energy. You're suggesting to view them as constants, and interpret the total energy as being a sum of pairwise terms. But they aren't constants. If you do that, you'll find that the "interaction energy" between two atoms will change if you move any other atom in the system. That's what I mean when I say it isn't a pairwise additive force.

[Siddharth Srinivasan]

Right. So I think the overall answer here is that its possible but complicated to get the entire force field to work with subsets, so I'll just shelve this for now. Thanks for your replies!