Ligand jumping around during GROMACS MD simulation after CHARMM-GUI setup – PBC issue or system prep

User discussions
1

GROMACS version: 2026.2
GROMACS modification: No

Hi everyone,

I’m running my first molecular dynamics simulation of a usnic acid derivative and would really appreciate some advice.

I started from the best docking pose obtained by AutoDock Vina.

The ligand PDB generated from docking contained generic atom names (e.g., C, C, C instead of C1, C2, C3, etc.).

My workflow was as follows:

  • Converted the ligand PDB to MOL2 using Open Babel.
  • Added hydrogens at physiological pH.
  • Assigned the correct formal charges in Avogadro.
  • Renamed all ligand atoms (C1, C2, O1, O2, etc.) in the MOL2 file exported from Avogadro.
  • Combined the ligand with the receptor to generate a complex PDB.
  • Renamed the ligand residue to UNK.
  • Uploaded the complex to CHARMM-GUI Solution Builder.
  • Generated the GROMACS input files.
  • Successfully completed energy minimization, NVT, and NPT equilibration.
  • Started a 100 ns production MD simulation.

While the production run is still in progress, I opened npt.gro together with the trajectory (md.xtc) in PyMOL/ChimeraX. I noticed that the ligand appears to jump around erratically, almost as if it is flying all over the simulation box.

My questions are:

  1. Could this simply be a periodic boundary condition (PBC) visualization issue, or does it more likely indicate an error during ligand preparation (such as atom renaming, charge assignment, or topology generation)?
  2. How can I determine whether the simulation itself is actually correct?
  3. Based on this observation, should I continue running the simulation, or is it better to stop now and investigate the problem?
  4. If I let the simulation finish the full 100 ns and obtain the final md.gro file, is it still possible to correct this issue afterward (for example, if it is only a PBC/imaging problem), or would I need to repeat the entire simulation?
  5. Does this type of ligand motion usually indicate that I made a mistake during system preparation, or can it still occur even when the simulation is completely valid?

Thank you very much for any suggestions or troubleshooting advice.

2

Hi @lhtrunggnhan

So by order:

  1. Yes, in principle, especially if you post-processed the trajectory, which doesn’t seem to be the case here. I suggest you to clean up the trajectory or, better, to use some tools like gmx distance to monitor the distance between the ligand and some residues in the binding site, so that you can see if the distance increases or not;

  2. This is a bit un-answerable as a question, what do you mean? At a force field level? System level?

  3. Generally it is better to run some kind of processing of the trajectories during the runs to monitor some quantities of interest so that you can kill simulations that produce garbage, as you will throw them away anyway;

  4. Yes if it is a PBC issue, no if it is not;

  5. This is a question that is impossible to answer in general. If a ligand leaves the binding site can be an equilibration problem, a force field problem, a binding pose problem, a protonation problem, etc. or maybe the ligand is just a weak binder. If you know experimentally that this is a very strong binder then I would start by questioning the pose and the force field choice, as well as the equilibration protocol.