How to avoid the molecules being too close together in multiple molecules MD simulation?

GROMACS version: 2021.4
GROMACS modification: No
I want to perform a protein-ligands syytem MD simulations. Unlike conventional MD simulations, the ligand were not put into the binding pocket, but I copied many ligand and put them around the protein to run MD.
But in the during MD, I found a fatal problem, that is, many ligand Molecules tend to clump together.

I tried to use [distance restraints] to control the spacing between small molecules, but currently the distance restraints module in Gromacs can only provide attractive force, not repulsive force.
Could there be other ways to prevent small molecules from attracting each other? Or is it possible to directly modify the potential form of distance restraints?

This isnt normal? if you keep the two molecules within their interaction limit it will attract or repel depends on its nature.
Are you trying to see which molecule bind to protein among the many? One way i can suggest is to keep them apart that it is beyong the electrostatic attractions. still i am not sure the validity of this approach.

I agree with you that the attraction between ligands is caused by the electrostatic contribution. But for now, I don’t expect them to be attracted to each other. Instead, I want them to remain free and not affect each other. So, as a concession, I can only try to add a repulsion between them.

There is no easy way to do this. For small ring systems, one can add a virtual site to the middle of the ring and an NBFIX-style repulsive interaction between the virtual sites. But for such large molecules, there is no straightforward way to implement such repulsion, because you have many rings (and thus would have intramolecular repulsion, which is going to lead to artificial behavior).

I would agree with the above assessment that such large, nonpolar molecules will aggregate like this. It’s consistent with physical reality.

Thank you so much for your suggestion to create virtual site. However, the interaction between virtual atoms and ligand atoms may change the configuration of the ligand, which is inconsistent with our previous purpose and may not be correctly introduced into the current study. A good practice is to limit the distance between the atoms in the two ligands to avoid the two being too close together. So I feel that modifying the potential form distance restraints should be a good choice.


I saw your post on the developers forum, so I’ll take a whack at it as well. These are just my opinions, but:

  1. If you made a massively concentrated solution of your ligand in real life it probably would “clump together” - as @jalemkul and @scinikhil already pointed out. I don’t think that Gromacs is wrong here!

  2. I don’t think that adding a repulsive term is going to work - you are going to end up spacing the ligand copies out through the unit cell. It seems like you really want all copies of the ligand to interact with the rest of the system but not with each other. This used to be possible in e.g. Xplor but not (as far as I know) in Gromacs.

  3. If you really want to do this you should probably run a very large number of simulations with single copies of the ligand positioned at different starting points and orientations all around the protein. This will be tedious!

  4. What are you trying to achieve here? Predict ligand binding pose? Find possible binding sites? If so there are many other programs already available for these purposes. Maybe not as powerful as Gromacs, but much faster!

1, I agree with you. But in real solvents, the aggregation of small molecules should not be so strong, and we all know that ligands will have diffusive behavior in solvents. However, by observing the animation of the MD simulation, I found that multiple ligands are basically in a state of aggregation.
2, I am not trying to completely isolate the interaction between the ligands, which is obviously not in line with the actual situation. What I want is to maintain proper spacing between small molecules.
3, This may be a feasible solution, but it is also a huge waste of resources.
4, I want to observe the process of small molecules into the binding pocket without any human interference . Similar to the actual situation, there should be many ligands surrounding a protein. At the same time, the simulation of multiple ligands and a single protein will undoubtedly increase the possibility of small molecules entering the pocket, which will give my project a chance of success.