Different Energies between Gromacs and OpenMM

[Ali Eltareb]

Hi,

I have been doing tests between Gromacs and OpenMM. I minimized a peptide using Gromacs (charmm36 forcefield). The energies that gromacs gave at the end of the minimization is

Code: Select all

   Energies (kJ/mol)
           Bond            U-B    Proper Dih.  Improper Dih.      CMAP Dih.
    1.67638e+01    2.25811e+01    1.28747e+02    2.69536e+00   -8.16862e-01
          LJ-14     Coulomb-14        LJ (SR)   Coulomb (SR)   Coul. recip.
    1.12933e+02    8.92176e+02   -4.19556e+01   -1.36340e+03    2.40821e+02
      Potential Pressure (bar)
    1.05456e+01    0.00000e+00

Steepest Descents converged to Fmax < 1000 in 4 steps
Potential Energy  =  1.0545593e+01

I took this minimized configuration and read it into OpenMM using GromacsGroFile and GromacsTopFile. I checked and OpenMM is reading the configuration correctly (its not adding or removing any bonds).

Code: Select all

Potential Energy:  10.345707164397743 kJ/mol
NonbondedForce -160.00702402028134 kJ/mol
HarmonicBondForce 17.842723235913645 kJ/mol
HarmonicAngleForce 21.15079444793464 kJ/mol
PeriodicTorsionForce 129.0225362249678 kJ/mol
CustomTorsionForce 3.157813891993246 kJ/mol
CMAPTorsionForce -0.8211366161302591 kJ/mol

The energies from OpenMM are slightly different than Gromacs (they do have the same order of magnitude). The differences are about 0.2-1 kJ/mol. I want to know why the HarmonicBondForce and HarmonicAngleForce from OpenMM doesn't add up to equal Gromacs Bond and U-B term. Is there something that I am missing? Could this be a platform issue?

Ali

[Peter Eastman]

I suspect you're evaluating them for a slightly different conformation. Here are some things to check.

• What precision are the coordinates specified to? If it's low precision, that could lead to slight differences in the energy.
• Do you change the coordinates in any way after you set them? For example, minimizing the energy again or enforcing constraints.
• Are the periodic box dimensions set correctly?

[Ali Eltareb]

You were right, the precision of the coordinates was too low (only 3 decimal places). I used a configuration from Gromacs with more precision and the energies are in great agreement. Thanks for the help.