[gmx-developers] more on electrostatic decoupling

[Berk Hess]

>
>
> I do object to the suggestion about grompp. One thing that's essential 
> for calculating  *standard* free energies of binding is applying 
> restraints between the ligand and the protein in the decoupled state 
> (both to ensure convergence and to get the standard volume into the 
> calculation somehow). See Boresch and Karplus, Gilson, Hermans, and 
> others for this. Anyway, the point is, it's necessary to restrain the 
> ligand to the protein to calculate absolute binding free energies. And 
> currently, GROMACS requires that restraints be *within* the same 
> "molecule"; that is, the ligand and protein must be defined in the 
> same molecule section in order to restrain the ligand to the protein. 
> But Berk's suggestion about grompp would make it so the ligand and the 
> protein would have to be in *different* molecules in order for grompp 
> to automatically fill in the parameters. If there isn't a workaround, 
> this would make it impossible to restrain the ligand to the protein.
>
> Presumably there is some other way to have grompp fill in the 
> parameters? Perhaps make it so grompp can be provided with an .ndx 
> file containing the atom list for the atoms to be decoupled (the ligand)?
>
Indeed one does not have to use molecules.
For an mdp option it would be even easier to implement one or more index 
groups.
All pairs and non-bonded interactions within each index group should be
converted with the scheme I described before.
Interaction of such an index group with the rest of the system or 
between two
index groups should be treated normally.
In that way you can also make multiple 'molecules' decouple simultaneously.
One should check that an index group does not extend over multiple 
molecules.

I can implement the three new pair interaction types.

Berk.