Why ligand leave the binding pocket during MD simulation?

GROMACS version: 2018.1
GROMACS modification: No
Forefield: charmm36-jul2020
Obtain ligand topology from CGenFF

Hello everyone,
I am working on a protein-ligand complex system, in 100ns md simulation ligand moves far away from pocket after 15ns. What could be the possible reasons and solutions?

That sounds more like a simple PBC correction issue. Make sure you’re properly re-imaging the trajectory. If the dissociation is real, then either the initial pose of the ligand was not right, or the topology is of poor quality such that stabilizing interactions are not maintained.

Thank you for response, I do re-imaging the trajectory with:
gmx trjconv -s md_0_1.tpr -f md_0_1.xtc -o md_0_1_noPBC.xtc -pbc mol -center
I use vina for docking and taking best poses. for two of my ligands, they not move out of pocket but for other ligands it happend. What should I do?

What are you using as the group for centering?

If the trajectory is properly re-imaged, and a ligand leaves when it shouldn’t, then the physical model of the ligand is probably incorrect, meaning some aspect(s) of its topology need refinement to properly stabilize its interactions with the protein.

For centering I select protein and when the ligand move out of pocket then I select non-water, which contain my protein and ligand but ligand came out of the pocket too.

Then you’ll have to assess whether the ligand topology is correct or if something else went wrong in preparing the simulation system.