[gmx-users] pressure scaling more than 1%

[van Bemmelen]

Hi Wang Qin,

It sounds to me like (part of) your structure immediately starts collapsing during equilibration. So I have a couple of suggestions:

1) It may help to take a look at the trajectory file from those first 500 steps, to see what's really going on physically.
2) Please make sure you also do your minimization run with the free energy paramters turned ON.
3) For equilibration runs, I'd suggest using Berendsen temperature coupling (with tau_t=0.2 or larger) instead of Nosé-Hoover, combined with the gen_vel mdp-option to generate initial velocities at 300K. Then, you may switch to Nosé-Hoover for your production run, with a tau_t of at least 0.5. Running with NH T-coupling on a system that's far from equilibrium can give very serious problems, resulting in various stange errors and crashes.

Apart from the above, I really don't see big problems in your mdp-file. Also, your strategy of taking the output structure from the previous lambda=0.9 (and re-minimize and re-equilibrate) sounds fine to me.

But maybe someone else with more experience has something to add to this. :-)

Grtz,
Jeroen


>From: "Wang Qin" <qwang86 at gmail.com>
>Subject: Re: [gmx-users] pressure scaling more than 1%
>To: "Discussion list for GROMACS users" <gmx-users at gromacs.org>
>Message-ID:
>	<564133630708090640m60599f02jd125374abcfb8cf5 at mail.gmail.com>
>Content-Type: text/plain; charset="iso-8859-1"
>
>The system is a protein binding with one ligand and the ligand 
>mutates to
>another one. The problem is found when lamda=1. First of all I 
>run a energy
>minimization using an initial structure from the product of 
>lamda=0.9. Then
>the problem happens when I run a equilibrium.
>Here is the mdp option:
>cpp                      =/usr/bin/cpp
>integrator               =
>md
>
>tinit                    = 0
>dt                       = 0.002
>nsteps                   = 50000
>nstcomm                  = 1
>nstxout                  = 0
>nstvout                  = 0
>nstfout                  = 0
>nstlog                   = 500
>nstenergy                =
>100
>
>nstxtcout                = 5000
>xtc-precision            = 1000
>nstlist                  = 1
>ns_type                  = grid
>
>
>Pcoupl                   = berendsen
>tau_p                    = 1.0
>compressibility          = 4.5e-05
>ref_p                    = 1.0
>
>
>tcoupl                   = nose-hoover
>tc_grps                  = system
>tau_t                    = 0.1
>ref_t                    = 300
>
>constraints              = none
>constraint-algorithm     =
>lincs
>
>shake-tol                = 0.0001
>lincs-order              = 12
>lincs-warnangle          =
>30
>
>coulombtype              =
>pme
>
>rcoulomb                 = 1.0
>
>epsilon-r                =
>1
>
>vdw-type                 =
>switch
>
>rvdw-switch              = 1.0
>rvdw                     = 1.3
>DispCorr                  =
>EnerPres
>
>fourierspacing           = 0.08
>
>fourier_nx               = 0
>fourier_ny               = 0
>fourier_nz               = 0
>
>
>pme_order                = 6
>ewald_rtol               = 1e-05
>epsilon_surface          = 0
>optimize_fft             = yes
>
>
>free_energy              = yes
>init_lambda              = 1.0
>delta_lambda             = 0
>sc_alpha                 =0.5
>sc-power                 =1.0
>sc-sigma                 = 0.3
>
>
>Thanks.
>
>
>On 8/8/07, Mark Abraham <Mark.Abraham at anu.edu.au> wrote:
>>
>> Wang Qin wrote:
>> > Dear all,
>> >           I am trying to do a FEP. In the equilibrium 
>step, it fails
>> > after about 500 steps with a warnning:  pressure scaling 
>more than 1%,
>> > mu: 1.06718 1.06718 1.06718. I searched the old discussion 
>about it and
>> > then try to increase the tau_p to 2 till 10. But it just 
>goes futher but
>> > still meets the problem at the end.
>> >           Any idea about it?
>>
>> Not unless you tell us how you prepared this system. Almost certainly
>> the initial structure is too far from the constant-P 
>equilibrium state.
>> Follow the kind of things that the tutorial material does, and also
>> check out this link
>> http://wiki.gromacs.org/index.php/Steps_to_Perform_a_Simulation
>>
>> Mark