[gmx-users] Clarity on TI free energy terms

[Hannes Loeffler]

On Wed, 1 Jun 2016 15:00:51 +0000
"Nash, Anthony" <a.nash at ucl.ac.uk> wrote:

> >  This also assumes that
> >you have vanishing atoms only.  If you have appearing atoms only you
> >would obviously have to revers the order, and when you have both you
> >will have to run with two mdp/tpr setups.
> 
> 
> With aspartic acid transforming into lysine on one polypeptide chain
> and then lysine transforming into aspartic acid polypeptide chain,
> all in the same system (keeps the charges the same and is a complete
> thermodynamic cycle), I will have both appearing and disappearing
> atoms. How do you mean ³to run with two mdp/tpr setups²? Is there an
> example you could give (which I am grateful for) or one in the manual?

Lambda paths are not selective in the sense that you could apply them
to only a subset of the system.  So if you have both disappearing and
appearing atoms you have to:
1) turn off the charges on the disappearing group (or alternatively all
charges to avoid a charged total system), q_off
2) turn off the vdW parameters for the disappearing group, vdw_off
3) turn on the vdW parameters for the appearing group, vdw_on
4) turn on the charges of the appearing group (or all charges), q_on

If you try to do this with Gromacs you will realise that the best you
can do is combine this into 2 mdp steps: 1) q_off and vdw_on/off and 2)
q_on.  Alternatively you can combine the vdw bit with 2) but that
doesn't make any difference.  Of course, if you could assume that a
combined Coulomb/vdW transformation will work, this would be a
non-issue...

Also, I wonder how pmx maps aspartate onto lysine.  I would think that
the gamma-carboxylate should not map onto the gamma-methylene but
rather the residue should be branched off at the C-beta and so
duplicate the rest of the residue.