[gmx-developers] Computing prot pot ener in the gmx code

[Mark Abraham]

David van der Spoel wrote:
> pascal.baillod at epfl.ch wrote:
> 
>> Hi!
>>
>> I am trying to get protein potential energies in order to carry out 
>> REMD simulations in function of the potential energy of a solute and 
>> not of the
>> entire system. I checked the tgroup.c file and found the following:
> 
> you will also have to treat PME in some way, which really is quite 
> complicated.

One horrible kludge to do this would be to run normal REMD for the 
period between exchange attempts, write output and exit. Now strip the 
solvent out of the output, construct vacuum input, do an energy 
calculation in vacuo (0 step simulation should do the job here). Now either

a) Do the REMD exchange attempts in a script using the energies you just 
calculated from their .edr files, and do any other housekeeping 
manually, reconstruct normal MD input files and iterate (you now want 
the REMD input files to never attempt exchanges, or every 10^20 steps or 
so :-) ), OR

b) Modify the gromacs REMD code to take the energies for attempting 
exchanges from either a file you constructed from the vacuo .edr files 
(or maybe those files themselves), and get gromacs to do the exchange at 
the first step of a new iteration of MD, without needing to construct 
new input files.

Which of those is more attractive depends whether b) is feasible (I've 
never tried to get gromacs to do an exchange attempt before 
integrating), whether you prefer hacking in a scripting language or 
gromacs and how much of the performance hit in a) is acceptable.

Mark