Open ZhuDaoScience opened 10 months ago
It would be helpful if you could provide a test case where OpenMM and GROMACS produce different energies or forces.
This is an example from the go-like Martini2 model, including gmx and openmm directories requested for tests. beta-hairpin_go_m2.zip
The results of this test case are as follows:
Gromacs and OpenMM energies do not match! Gromacs: -238.376 OpenMM: -13018.091 Delta: 12779.716 Relative: 53.612
Gromacs and OpenMM forces do not match! Values differ at 39 of 114 coordinates.
First differing position: (0, 0)
Gromacs: -87.001100 285.807000 -171.122000
OpenMM: 150058.392542 -212920.652010 219791.880078
Difference: 150145.393642 213206.459010 219963.002078
Largest violation is at: (0, 2).
Gromacs: -87.001100 285.807000 -171.122000
OpenMM: 150058.392542 -212920.652010 219791.880078
Difference: 150145.393642 213206.459010 219963.002078
Error: The force of Category [ Pairs ] is error! In Gromacs, the interactions in [ Pairs ] are plain Lennard-Jones which are not affected by the non-bonded interaction treatment and potential modifiers, which means that there is no cutoff or modifier in the force item. However, in the scripts of Martini-Openmm, the interactions in [ pairs ] are treated as the same as normal L-J interactions. I have tested that it is error in the codes.