[gmx-users] free energy and soft core

[Frauke Meyer]

Dear Berk,

thanks for your answer.
No matter what the structure is and if there are bad contacts, the
settings
free_energy              = yes
init_lambda              = 0
delta_lambda             = 0

should result in an MD that does *not* depend on the soft-core
potential, right? The reason is that the full potential is just the
potential of topology A, since  V = (1-lambda)*V_A + lambda*V_B. And
V_A is equal to the hard-core potential for lambda=0.

I find a stable MD for 
sc-alpha                 = 0
whereas mdrun crashes after 200 steps with LINCS warnings for
sc-alpha                 = 1.5


dG/dl is different for both, which is alright.
                alpha=0        alpha=1.5
                @TYPE xy        @TYPE xy
                0 173.378       0 173.378
                0.002 170.874   0.002 -7.056
                0.004 168.351   0.004 -7.9578
                0.006 166.028   0.006 -8.24761
                ....


But should not the potential energy itself be identical for lambda=0?
What am I doing wrong?

Regards,
Frauke

PS: grompp gives:
> WARNING 2 [file fep.mdp, line unknown]:
>   You are using lattice sum electrostatics with free energy 
>   integration.
>   This might give wrong results, since the lattice contribution to the
>   free energy not calculated.

Should I better not use PME with FEP then?

Frauke


On Mon, 2004-08-09 at 16:08, Berk Hess wrote:
> 
> 
> 
> >From: Frauke Meyer <fmeyer at iris3.simm.ac.cn>
> >Reply-To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> >To: gmx-users at gromacs.org
> >Subject: [gmx-users] free energy and soft core
> >Date: 06 Aug 2004 10:57:52 +0800
> >
> >Hi all,
> >
> >for the perturbation of a ligand (A) to dummy atoms (B) I want to use
> >the
> >soft core potential. With
> >
> >free_energy              = yes
> >init_lambda              = 0
> >delta_lambda             = 0.000001
> >sc-alpha                 = 1.5
> >sc-sigma                 = 0.3
> >
> >I get LINCS warnings after the first few steps and mdrun crashes,
> >whereas
> >with
> >
> >sc-alpha                 = 0
> >sc-sigma                 = 0.3
> >
> >the whole perturbation runs fine (though with the expected singularities
> >for lambda near to 1).
> >
> >Surprisingly, with larger sc-alpha (eg. 10) LINCS warnings are fewer and
> >stop after some steps, and the perturbation terminates normally.
> >What might the problem here be? Topology B (dummies covalently boudn to
> >each other) anyway has no dispersion. I expected less warnings and
> >similar
> >results to alpha=0 with *lowering* alpha, but did not find that.
> >
> >Another surprise was that the dynamics depend on sc-sigma: Higher
> >sc-sigma
> >has the same effects a higher sc-alpha. But should not sigma_A be
> >calculated brom C6_A and C12_A, and sigma_B not play a role, because
> >C6_B = C12_B = 0 anyway, so that sc-alpha is not used for the distortion
> >of LJ?
> 
> This is correct.
> But sc-alpha is also used for particles that only have charge and no LJ
> interactions. Could it be that you have such kind of atoms in your such,
> for instance hydrogens?
> 
> It is not surprising that with a larger sc-alpha everything goes fine.
> A larger sc-alpha makes a larger soft core and thus the whole
> potential gets smoother.
> 
> But the difference between 0 and 1.5 is quite strange.
> Also a few steps into the simulation lambda should be almost 0
> (or am I wrong?) and thus there should be almost no soft-core
> effect present yet. Therefore I would expect the same behavior
> independent of the sc-alpha value.
> 
> The problem could just be a bad starting structure, too close
> atoms contacts or something like that.
> 
> Berk.
> 
> _________________________________________________________________
> Play online games with your friends with MSN Messenger 
> http://messenger.msn.nl/
> 
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
>