[gmx-users] Group NS scheme parameter for 2016.3 version

Mark Abraham mark.j.abraham at gmail.com
Tue Sep 26 13:44:56 CEST 2017


Hi,

On Tue, Sep 26, 2017 at 4:02 AM Du, Yu <duyu at sioc.ac.cn> wrote:

> > -----Original Messages-----
> > From: "Mark Abraham" <mark.j.abraham at gmail.com>
> > Sent Time: 2017-09-26 06:38:40 (Tuesday)
> > To: "Discussion list for GROMACS users" <gmx-users at gromacs.org>,
> "Discussion list for GROMACS users" <
> gromacs.org_gmx-users at maillist.sys.kth.se>
> > Cc:
> > Subject: Re: [gmx-users] Group NS scheme parameter for 2016.3 version
> >
> > Hi,
> >
> > On Mon, 25 Sep 2017 14:11 Du, Yu <duyu at sioc.ac.cn> wrote:
> >
> > > Dear GMX Users,
> > >
> > >
> > > Indeed making tabulized potential between groups with Verlet is not
> > > trivial as also discussed in OpenMM issue #1765.
> > >
> > > I'm using gmx 2016.3. If I want use the group NS scheme as precise as
> > > possible, what parameters in mdp file should I set and what are their
> > > recommended values for protein and ligand system.
> > >
> >
> > It's straightforward to implement a tabulated interaction kernel, but
> > rather less clear how best to let the user describe to mdrun how they
> want
> > the calculation to work, eg whether interaction types should be governed
> by
> > atom numbers, or types, or names, and how that should be expressed in the
> > topology. Suggestions for problem types people want to be able to handle
> > are most welcome.
> >
>
> I want to decompose the interaction between protein and ligand, and only
> between them, not others.
> So I think "vdwtype=user, energygrps = Protein Ligand, energygrp-table =
> Protein Ligand and table.xvg table_Protein_ligand.xvg" is the only solution
> in gromacs to achieve it.
>

Yes. So in future, some method that would select the regions to use the
different tabulated interactions based on atom number (ie index groups)
would work for you.

> How to set up the nonbonded scheme for a given force field varies with the
> > force field, either based on how it was parameterized or been shown to
> work
> > in practice. This is standard practice. A major defect of the group
> scheme
> > is that it is inefficient with a buffered list, but that is your tradeoff
> > to make.
> >
>
> Yes, this is my major confusion. Amber FF don't use charge group not like
> GROMOS FF. I need to work with Amber FF, so is it proper to use
> amber99sb-ildn and gaff with group cut-off scheme?
>

Using charge groups of size 1 is the same as not using charge groups, which
is how the AMBER force fields were already implemented, long before the
Verlet scheme was implemented. You should also be reading other people's
work that used GROMACS and the AMBER force fields...


> And now the manual and user guide is dominated by Verlet cut-off scheme,
> where can I find the guide of using group cut-off scheme? I can't tell
> which options is for group scheme in the Molecular dynamics parameters
> (.mdp).
>

Nothing's been removed, and nearly everything works the same way.
http://manual.gromacs.org/documentation/2016.4/user-guide/mdp-options.html#neighbor-searching
and
the coulomb and vdw sections are the only bits that matter, and they note
the points of difference.


> The following is what I found in
> regressiontests-2016.3/kernel/nb_kernel_ElecCSTab_VdwCSTab_GeomW4W4/grompp.mdp
> ; NEIGHBORSEARCHING PARAMETERS
> ; cut-off scheme (group: using charge groups, Verlet: particle based
> cut-offs)
> cutoff-scheme            = Group
> ; nblist update frequency
> nstlist                  = 1
> ; ns algorithm (simple or grid)
> ns-type                  = Grid
> ; Periodic boundary conditions: xyz, no, xy
> pbc                      = xyz
> periodic-molecules       = no
> ; Allowed energy drift due to the Verlet buffer in kJ/mol/ps per atom,
> ; a value of -1 means: use rlist
> verlet-buffer-drift      = 0.005
> ; nblist cut-off
> rlist                    = 0.99
> nstcalclr                = -1
>
>
> Although it use group cut-off scheme, there is also a verlet-buffer-drift,
> which doesn't show in the 2016.3 version .mdp option.
>

Don't use a test input for designing a real simulation. Those are designed
to do very specific things, including hit corner cases that we think nobody
should use. And that parameter is an old name that is not a problem when
the group scheme is chosen.

You should find some literature that shows the force field and software
working together, and test it yourself also. Unfortunately I know of nobody
who's stepped up and published "this is how you should run *these kinds* of
simulations" for AMBER ff in GROMACS.

So, lots of thanks for anyone's advice for setting group scheme.
>
> > Given that you are modifying the force field to add different
> interactions,
> > you will anyway have to show that the modified form of the force field
> > works suitably. Even if the Verlet scheme had support for tabulated
> > interactions, you would still be limited to having only one set of
> > nonbonded parameters per atom type. Is your mix of interaction types
> > feasible to use with such a restriction?
> >
>
> I'm really looking forward to the combination of Verlet cut-off scheme
> with the tabulated potential. :)
>

That wasn't the question. Can your ligand atoms have *one* set of vdw
parameters that can interact with solvent, ligand and protein atoms
according to different functional forms? People sometimes ask for such
things without thinking through whether their model makes sense (and if it
does, whether they would require schizophrenic atoms).

Mark

> Mark
> >
> > >
> > --
> > Gromacs Users mailing list
> >
> > * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>


More information about the gromacs.org_gmx-users mailing list