Amber19SB and Amber14SB in old versions of Gromacs

User discussions
1

Dear all,

In the 2026 release, there are new amber force fields that are available (ff14SB and ff19SB). I am wondering if they could work on old versions of Gromacs? The CMAP is available since Gromacs 2025 if I am correct: is it enough for any of these two force fields to work smoothly with Gromacs 2025? The reason I am asking is because I would like to run HREX (REST2) simulations with these force fields, and plumed cannot patch (yet) Gromacs 2026, but can do it for Gromacs 2025.0.

Thanks for the answer

Nicolas

2

No, new functionality is needed to assign the proper CMAPs the sidechains of residues.

3

Thanks. It makes sense, but I wanted to check.

4

We have builtin PLUMED support. Patching should not be necessary anymore.

5

HREX is not supported by the builtin PLUMED.

6

I see. Indeed, then the answer is no. 2025 also lacks LEaP-like dihedral reordering functionality and support for per-water model ions.

7

@hess Since Amber14SB doesn’t use CMAP, what are your recommandations/thoughts regarding using the official port on older versions ? I have tried, and simulations can start with Gromacs 2022.5 using the Amber14SB from Gromacs 2026.0 (and as expected it fails with Amber19SB).

8

Ah, then I mixed up things. I suppose Amber14SB should work in older GROMACS versions then.

@vedran can you comment?

9

Assuming that the ions are not used at all and that there is some tolerance for dihedral energy mismatch with Amber proper, sure. I am not sure what is the case with water models, but I expect that they can be specified in the command line and should work fine.

Overall, I wouldn’t recommend it, but it should be fine for experimentation until 2026 is supported by PLUMED patching.

10

@vedran I am not sure to understand what you mean about the ions. Straight out of the box, the combinaison Gromacs2022.5/Amber14SB failed. But then I did first the following steps : pdb2gmx / editconf / solvate / genion. Then I opened the .top and manually added : #include “amber14sb.ff/ions_tip3p.itp”. Then, grompp/mdrun for energy minimization then MD, and everything ran smoothly. Do you see a caveat there ?

Regarding the mismatch with dihedrals, I think I can live with that.

Thanks

11

Assuming your water model is indeed tip3p, that should be fine.

I can’t recommend what you are doing because we have only tested ff14SB and ff19SB with the code that later became GROMACS 2026, but I can’t think of other possible issues at the moment either. In any case, please proceed with caution.

12

Thanks for the answer. I understand your point. If I want to take the risk, what would you recommend to validate what I propose? Running two NVE simulations (one which each version of Gromacs) starting from the same .tpr, and comparing the potential energies along the trajectories? Or extracting 1000 snapshots and comparing single point energies with the two versions?

13

That would be the first order of validation, yes, and it can be done reasonably fast. Anything stochastic is likely to require fairly long simulations.