Free energy perturbation - mdp parameters

User discussions
,
1

GROMACS version: 2019
GROMACS modification: No

Hi GROMACS users,
I’m new to free energy perturbation and I would greatly appreciate your input. In a nut shell, I wish to alchemically transform an aspartate in a protein’s binding site from de-protonated to protonated sate (the ligand - an ion - does not changes). For this purpose I produced a hybrid protein structure and topology using the pmx server.

First, as this is quite a large system (protein+membrane+solvent), is it legit to equilibrate the system once and then replicate it with the different lambda values in the mdp file, dropping the first x-ns of production for the final analysis?

Second, I would appreciate some input on the following mdp options (a complete mdp file is also attached):

couple-moltype, couple-lambda0 and couple-lambda1: I left these options empty. The gromacs manual implies one cannot use both hybrid topologies and these mdp file options. Is this the case?

couple-intramol = yes: becauese the molecular structure changes (protonation state).

mass_lambdas and bonded-lambdas: as I’m adding a hydrogen atom to my aspartate residue I’m assuming I should set these two options alongside the coul-lambdas and vdw-lambdas. I intend to first decouple the coulombic interactions and then vdw. When should the mass and bonded interactions be decoupled? Parallel to vdw?

sc-coul = no: In equilibrium alchemical free energy calculation only the van der Waals interactions need to be soft-cored. Right?

Any help would be much appreciated!

Cheers,
Gal.

fep.mdp (4.2 KB)

2

The couple parameters are for solvation free-energies, you don’t need those.

Avoid changing masses if you can, then you don’t need mass-lambdas either.

Are you planning to do the whole changes in a single lambda path? Then you need to couple the bondeds and sc-coul=yes, as you are growing a hydrogen, I assume with charge. Note that in thi, s case it is simpler to do everything with a single lambda setupwithout the lambda for interaction types.

3

Hi hess, thanks for replying!

I am planning on doing the whole changes in a single lambda path.
I’m not sure I understand what you mean by “do everything with a single lambda”. Do you mean using fep-lambdas instead of coul/vdw etc? or is it something else?

Cheers,
Gal.