GROMACS version: 2022.3
GROMACS modification: Yes/No
I am trying to run a molecular dynamics simulation of a customized 756-nt RNA model generated using AlphaFold3, and I have encountered multiple issues during topology generation and energy minimization. First, after cleaning the PDB, grompp produced the error “OP3 in residue A 1 was not found in rtp entry RA5 with 31 atoms while sorting atoms,” which I traced to the free 5′-terminal phosphate; the -ter option did not resolve it, so I manually removed the extra phosphate. Next, during em.tpr generation, I received a warning about excluded atoms being separated by 7.9 nm, along with a very large PME grid; I proceeded with -maxwarn 1. When running minimization, GROMACS reported “There is no domain decomposition for 40 ranks that is compatible with the given box,” which I addressed by using -ntmpi 1 as suggested in the GROMACS 2025.4 documentation. Minimization then started but failed at step 42 with “One or more water molecules can not be settled… Epot = inf, Fmax = 6.6e+05,” indicating severe bad contacts within the solvated system. I would be grateful for guidance on best practices for preparing and simulating very large RNA systems in GROMACS, including how to properly handle terminal phosphate groups, how to fix such early-stage water-settling/bad-contact issues, and whether it is possible to add 5′/3′ caps to RNA residues in silico within standard workflows.
It’s possible that your RNA molecule has a discontinuity - maybe it’s two chains, or one chain with a gap, that Gromacs internally sees as a single molecule. You’re right that you typically have to remove the 5’ phosphate to correctly pass through pdb2gmx.
Try minimizing the molecule in vacuum (just as a validation), this will obviously distort it so don’t use it later on, but from the final structure you will clearly see if there are any spurious bonds that shouldn’t be there. Otherwise, these water placement errors might be obscuring more serious issues.