[gmx-developers] Coming on to help in GROMACS development

[Berk Hess]

Michael Shirts wrote:

>Sorry for the delayed response!
>
>  
>
>>What do you mean with decoupling?
>>    
>>
>
>What I meant was that if you use a free energy perterbation to a
>molecule in a larger system, at least in GROMACS 3.1.4, the
>intramolecular terms would also be lambda dependent.  If you were
>interested in the solvation free energy of octanol, for example, this
>wouldn't be good, since one end of the molecule wouldn't be able to
>"feel" the other end when lambda became zero.  You could do another
>simulation turning off all the terms in the vapor state, but that's
>becomes difficult and annoying.
>
>  
>
This depends a lot on the system I guess.
For many system you do want to turn of the intramolecular non-bonded
terms as the ensemble of the molecule would become very different
in the vapor state, whereas a molecule without non-bonded interactions
might have an ensemble closer to the solvated state.

>It's actually possible to work around this for the Lennard-Jones
>potential, by defining individual ij terms, but not for the Columb
>part.  By making slightly different neighborlists, however, it's
>fairly easy to do, and is what I implemented.
>  
>
I guess that in put_in_list you would just have to check if both atoms
are perturbed instead of only one in the pair.
I have also made some changes put_in_list, especially for free-energy 
simulations

>  
>
>>I have recently implemented free-energy support for Ewald and PME.
>>Also I have added reaction-field corrections for the exclusions,
>>with makes free-energy completey correct with RF.
>>    
>>
>
>I assume these code changes are you in the CVS version, correct?
>
>  
>
Yes.

>>What kind of restraints?
>>Position restraints now also support free-energy?
>>    
>>
> 
>Yes, position restraints.   For example, in the ligand binding paper
>we submitted, we use a harmonic restraint in the unbound state to make
>the sampling of that state simpler, and turn this restraint off as a
>function of labmda.  We can then correct for the restrained state
>analytically.
>  
>
I accidentally put a question mark at the last line.
I meant that I have implemented this and that it is in CVS.

Berk.