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

[Mark Abraham]

Hi,

On Wed, 24 Jul 2019 at 21:33, Kenneth Huang <khuang8 at student.gsu.edu> wrote:

> Hi Mark,
>
>
> Thanks for the reply- I guessed a number of things had changed in the last
> decade, but wasn't sure if it was even salvageable as a comparison point
> anymore.
>
>
> Out of curiosity, could you elaborate on what sort of measurements I could
> make to scale the difference? Since I don’t have a run with the older
> version, wouldn’t anything I measure reflect the current version? I imagine
> it can't be as simple as adjusting the cutoffs in the electrostatics.


The cutoff determines the shift, which is the same quantity on *each*
interaction. So as I said, you could measure the shift applicable in your
case (from the value of the Coulomb interaction at the cutoff), and
estimate the number of interactions of each kind of charge-charge pair to
which it applies, to see if the difference is plausible. But I don't see
that as a useful way of assessing whether the modern code can reproduce the
intended observables, which don't relate to the absolute energies.

Mark