[gmx-users] topology files for ligands in MD Simulation under OPLS AA force field

[Michael Brunsteiner]

mircial at sjtu.edu.cn mircial at sjtu.edu.cn wrote

> I am using GROMACS package to do molecular dynamics
> simulations under OPLS_AA force field. I encounter
> some problems when preparing the topology files
> of small molecules (ligands).My questions are as follows: 
> 1, how to chose the atom type of each atom from the ligands? 
amber + GAFF would be easier, here the ligand parameterization
is pretty straight forward (there's probably a HOWTO)

with OPLS you want to work with homology, and if you don't need
any high throughput you can do it manually ... look up 
groups in amino acids that are similar to your the components
of your ligands, and choose the atomtypes accordingly.

> 2, how to define the charges of each atom? I know that, when
> the atom type is defined, there will be a corresponding charge
> to this atom type. Does it safe to use this charge in the
> simulations on the ligands (since these charges are designed
> for amino acids)?

yes, OPLS was mostly used for peptides, but AFAIK, the
original parameterization in Jorgensens group started with generic
organic molecules, there's certainly papers from this group that
present parameters for non-peptide molecules, e.g., various
hetero cycles ... (I am not sure whether the gmx topology folder
only contains the part of OPLS relevant for peptides, or all of it,
if not ask someone from the Jorgenson group, they might give you
the original parameter files)
However, if you have some generic organic molecules and assign
OPLS atom types there is no guarantee that the charges add up to
zero (for a neutral molecule) ... what i did before, with
reasonable success, is combine EPSD (SCF or AM1BCC) charges with
OPLS parameters...

In any case, as has been pointed out before many times in this list
mixing different FFs, or using ad-hoc values for some parameters, is 
dangerous, and you really want to test your molecules before using 
them in any extensive calculations. 
A good start might be to compare structures that were minimized
with your FF to the same structure minimized at the DFT or MP2 level
(HF-SCF might do as well depending on the molecule) ... but that's
probably not enough ...

> 3,When the atom type and the charge of atom are defined,
> how to prepare the file in the GROMACS format? Does there easier
> method to prepare such files than manually? 
As for the bonded interactions, in some cases pdb2gmx and
grompp can figure them out by themselves. You'd assign atom- 
(thereby bond-) types, and the bonds, then make a new entry
in the rtp file, including only the atoms and bonds sections. you 
choose a new 3 letter residue name, and starting from a PDB
file of the ligand (with this residue name, and the same atom order
and the same atom names as in the rtp entry) use pdb2gmx/grompp.
However often enough, especially for proper and improper dihedrals
you have to define them manually (again by anlogy) 
for anything but the simplest molecules its a painstaking task ...
so you either take the time, or you use amber (or you talk to 
Schrodinger, they will have some script for assigning OPLS parameters
but that probably wont work with Gromacs)

good luck!
mic