[gmx-users] Fatal Equilibration Errors

[Nancy]

The structure in the original PDB file of ethylene glycol was in the gauche
conformation.  When I minimised the solvated system, the molecule remained
in the gauche conformation.  However, during equilibration of the system,
the conformation of the molecule changed to trans within less than one
picosecond.  It remained in the trans conformation for the rest of the
equilibration.


Thanks,
Nancy



On Sun, Aug 9, 2009 at 5:49 PM, Nagy, Peter I. <PNAGY at utnet.utoledo.edu>wrote:

>
> Dear Nancy,
>
>   It was useful to find charges for the solute... I am not a GROMACS user,
> so I do not know what checks have to be done for a non-protein or non-DNA
> solute. Anyway, this case helped develop data management for a non-standard
> solute.
>   It is normal to use united atom model for CH, CH2 and CH3 in GROMACS. As
> far as the charges are concerned, it is always a matter of speculation.
> Nobody knows what charges are "correct". Only a quantum MD (for the whole
> system) helps, where you need not derive charges since you do not use
> empirical pair potentials. Charges of -0.574 and 0.398 for the OH atoms are
> reasonable, although somewhat larger absolute values could also be used.
> Nonetheless, they already triggered the solute-solvent H-bonds, which was
> the Nr. 1 problem.
>   I do not know what conformation you derived from the web. Was it
> ORIGINALLY trans or gauche? As I wrote, the experiment found only 12% trans
> in aqueous solution. Try to start the structure from the so-called tGg'
> conformation or some other gauche(you may find them e.g. in our papers
> JACS,1991, 113, 6719; 1992, 114, 4752 and a complete set of conformations in
> the paper of Cramer and Truhlar in the JACS about 1995, I do not have the
> reference handy). If the system stays in gauche, then the simulation has
> been short starting from trans and/or the OCCO torsion barrier is too high
> for equilibration. I would start from the tGg' and would conduct at least
> 500 ps long simulation, check if trans appears and what is the gauche/trans
> ratio. If you do not find a trans starting from gauche or no gauche starting
> from trans, then I would reduce the OCCO torisonal. Also check whether the
> intramolcular hydrogen bonds remain maintained in the gauche structures.
> Hopefully these above help.
>   The ratio is a sensitive problem, and could depend on the applied charge
> sets, too. Thus you may need to conduct a more complicated test finding
> combination(s) of charges and torsion parameters, which produce good
> results. Maybe there are different combinations working equally well. I did
> not meet this problem when I performed the free energy perturbation
> calculations via Monte Carlo simulations with rigid solute geometries.
>
> Best regards
> Peter Nagy
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20090810/b20d74b4/attachment.html>