[Fwd: Re: [gmx-users] QMMM free energy]

[David Mobley]

Gerrit and David,

David wrote:

>But wouldn't BAR work just fine?

And On 5/31/07, Gerrit Groenhof <ggroenh at gwdg.de> wrote:
> Hi Fiske,
>
>  From what I understand, you want to compute the free energy for
> changing one or more atoms in the QM subsystem. I never considered doing
> this because I do not expect to get sufficient sampling. But in my
> opinion this should be possible.
>
> In case only atoms in the QM subsystem change it is straightforward. But
> I might overlook something important (maybe someone can comment on this
> if I make a mistake).
>
> I think you should modify the code, so that there are two QM/MM
> computations per step. Once in the A state and once in the B state,
> which gives VA and VB. From that you can compute
> V(lambda) = lamda VB + (1-lambda) VA. And dv/dl = VB-VA. This is then
> the QM and coulombic QM/MM contribution to dv/dlambda. The lennerd jones
> should work automatically.
>
> But this works only if you are not modifying the charges of the MM
> atoms. Then things get more messy.
>
> However, there is a problem with neighbour searching. Right now, the
> QM/MM neighbour list is constructed only in the
> if (!bFreeEnergy)
> branch of the put_in_list function. I think this is why mdrun crashes.
> the qmmm interface expects to find MM neighbours, but the QMMM nblists
> are not created.
> You can copy the if(bQMMM) branch into the free energy branch.
>
> Best,
>
> Gerrit

Both of you guys are thinking linear mixing of the two Hamiltonia,
where one can just evaluate the energies of the end states (lambda=0
and lambda=1) and then combine to get the potential (or dV/dlambda) at
the intermediate lambda values. That's fine as far as it goes -- but
it won't work for doing insertion/deletion of atoms, where one needs
to use something like soft core potentials.

Unless I'm missing something obvious, it's not at all clear to me how
one would do insertion/deletion of atoms using soft core potentials
with QM/MM. Though I don't know much about QM/MM, I do know it
involves doing some sort of QM calculation -- and I don't get how that
could possibly work with something like soft core potentials, since
the QM calculation must use an atom type. If you're deleting an atom,
you want, at lambda=0.5, say, an atom that is halfway between a dummy
atom and what it originally was (say, a carbon). I have no idea how
you'd describe that quantum mechanically.

And as far as whether BAR would save the day...  Nope. You need to
have the potential energy of your snapshots evaluated at different
lambda values as input for BAR, which means you have to figure out how
to evaluate a QM energy for an atom that is halfway between a dummy
atom and a carbon atom. I don't get how this would work.

Maybe someone has worked this out already -- again, I'd look at what
Wei Yang, or anyone else doing QM/MM free energy calculations, is
doing. But it's definitely not something that's immediately obvious.

David

>
>
>
>
> >
> > Subject:
> > Re: [gmx-users] QMMM free energy
> > From:
> > Erik Lindahl <lindahl at cbr.su.se>
> > Date:
> > Wed, 30 May 2007 17:34:59 +0200
> > To:
> > Discussion list for GROMACS users <gmx-users at gromacs.org>
> >
> > To:
> > Discussion list for GROMACS users <gmx-users at gromacs.org>
> >
> >
> > Hi,
> >
> > I'm not a QM expert myself (Gerrit, comments?),  but I'm pretty sure
> > the main reason is that you cannot either mutate atom types or
> > decouple interactions gradually in the QM part of the system. Once
> > this is solved for a pure QM system it should be straightforward to
> > implement the QM/MM interface of it.
> >
> > Cheers,
> >
> > Erik
> >
> > On May 30, 2007, at 5:30 PM, Robert fiske wrote:
> >
> >>> With the original QM/MM implementation, free energy is currently not
> >>> possible (yet). From what you report, I guess that also with the
> >>> CPMD interface it does not work.
> >>
> >> The CPMD interface is mostly plugged into the same setup as the other
> >> QM interfaces, I was able to bypass the error by copying modifying
> >> the free energy else statement to bypass the vdw calculations for the
> >> case of bQMMM true (and generally try to have that code match the
> >> bQMMM case of non free energy runs).  But since my background is
> >> Computer Engineering and not chemistry (Chemistry 101 is as far as I
> >> got in formal training)  I don't really know what should be done (I
> >> assume since as you said it's not implemented yet although I don't
> >> get a run time error that the results will be off).
> >>
> >> Are QM/MM free energy calculations substantially different than MM
> >> calculations, or is it just not implemented because no one has needed
> >> to do QM/MM with free energy yet?
> >>
> >>
> >> fiske
> >> Gogonea Group
> >> Cleveland State University
> >>
> >> _________________________________________________________________
> >> More photos, more messages, more storage—get 2GB with Windows Live
> >> Hotmail.
> >> http://imagine-windowslive.com/hotmail/?locale=en-us&ocid=TXT_TAGHM_migration_HM_mini_2G_0507
> >>
> >>
> >> _______________________________________________
> >> gmx-users mailing list    gmx-users at gromacs.org
> >> http://www.gromacs.org/mailman/listinfo/gmx-users
> >> Please search the archive at http://www.gromacs.org/search before
> >> posting!
> >> Please don't post (un)subscribe requests to the list. Use the www
> >> interface or send it to gmx-users-request at gromacs.org.
> >> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> >
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > Please search the archive at http://www.gromacs.org/search before
> > posting!
> > Please don't post (un)subscribe requests to the list. Use thewww
> > interface or send it to gmx-users-request at gromacs.org.
> > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
> >
> > ------------------------------------------------------------------------
> >
> > Subject:
> > Re: [gmx-users] QMMM free energy
> > From:
> > "Robert fiske" <rfiske_ at hotmail.com>
> > Date:
> > Wed, 30 May 2007 12:00:29 -0400
> > To:
> > gmx-users at gromacs.org
> >
> > To:
> > gmx-users at gromacs.org
> >
> >
> >
> >> I'm not a QM expert myself (Gerrit, comments?),  but I'm pretty sure
> >> the main reason is that you cannot either mutate atom types or
> >> decouple interactions gradually in the QM part of the system. Once
> >> this is solved for a pure QM system it should be straightforward to
> >> implement the QM/MM interface of it.
> >
> > One of the main things we plan to do with the free energy code (once I
> > can confirm I'm doing things right) is to modify the charge of the
> > quantum system for the lambda values, so atom types don't need to be
> > modified for this I believe, although I see how that could complicate
> > other types of free energy calculations.  As for decoupling
> > interactions I don't know what impact that has for changing of charge.
> >
> > For now, are there any major complications if all we want to do is
> > modify the charge from state A to state B (assuming the interaction
> > problems you mentioned aren't a problem)
> >
> >
> > fiske
> >
> > _________________________________________________________________
> > Make every IM count. Download Messenger and join the i'm Initiative
> > now. It's free. http://im.live.com/messenger/im/home/?source=TAGHM_MAY07
> >
> > _______________________________________________
> > gmx-users mailing list    gmx-users at gromacs.org
> > http://www.gromacs.org/mailman/listinfo/gmx-users
> > Please search the archive at http://www.gromacs.org/search before
> > posting!
> > Please don't post (un)subscribe requests to the list. Use the www
> > interface or send it to gmx-users-request at gromacs.org.
> > Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>
> _______________________________________________
> gmx-users mailing list    gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php
>