[gmx-users] Advice for free energy perturbation calculations

[Raphaël Robidas]

I am somewhat new to GROMACS and molecular dynamics. I have been
experimenting with different methods of doing free energy calculations. In
particular, I am trying to integrate the energy difference between two
enantiomers when bound to a chiral ligand. GROMACS has many options and I
am not sure which one to use. I have some experimental data to compare to,
but they are not directly linked to the free energy difference; thus I am
not sure if I am doing the free energy calculations wrong, or if my model
(e.g. the hypotheses I make to link the free energy difference to the
observed quantity) is incorrect.

Here are a couple more specific questions related to this:

1. Many tutorials only use the soft-core option, but I don't know if it's
useful in my case. GROMACS itself says that this option does not induce
errors, but requires more sampling. From my understanding, it is only
useful when there is total annihilation of a molecule, right?

2. The integration can be done with discrete steps of lambda, or by a
continuous variation of lambda (using delta-lambda = ...) The continuous
integration seemed superior to me, but most tutorials use discrete
integration. MBAR is also only available for discrete integration (as far
as I know). What is the appropriate situation for each of those methods?

3. Another method for the integration would be using the expanded ensemble.
I tried using it, but ran into some errors which lead me to another email
in this mailing list saying that the option was pretty experimental. Is
this still true/should I use this option?

Thank you very much!