[gmx-users] Variation in free energy between GROMACS versions?

[Kenneth Huang]

Hi all,

I'm trying to implement a variation of  the simulated tempering procedure as described in a 2007 paper (Choosing weights for simulated tempering, Sanghyun Park and Vijay S. Pande, Phys. Rev. E 76, 016703) on GROMACS 2018, and am trying to verify that my implementation is correct compared to the system in the paper (Ala10). While I've managed to match a majority of the parameters that are described in the paper, the energy values I'm getting are all shifted about 3,000 kJ/mol relative to the ones in the paper.

I'm not familiar enough with how mdrun itself has changed within GROMACS, and don't expect to fully be able to reproduce the results of the paper, so I was wondering if that large of a difference is expected? Considering that I'm using as many of the same settings as in the paper (ie, water model, forcefield, number of waters, thermostat, etc), is it unusual to see that magnitude of a change for a relatively simple system like Ala10 between an older and newer version of GROMACS? Or is this something that's expected, and I shouldn't worry to much about it?

Best,

Kenneth