Re: [AMBER] PMEMD does not support intermolecular PRFs

From: Jason Swails <jason.swails.gmail.com>
Date: Fri, 17 Jun 2011 00:37:56 -0600

On Thu, Jun 16, 2011 at 8:05 PM, Maura Catherine Mooney <mmooney05.qub.ac.uk
> wrote:

> Hi Jason,
>
> I appreciate the concise reply and I'm not intending to drag this out, but,
> I had another look at a trajectory in vmd, and as far as I can see there is
> only one GDP residue. I choose 'resname GDP'. I looked at a few attempted
> trajectories and haven;t observed anything other than one GDP residue name,
> which does correspond to the GDP molecule (in VMD). Is there something else
> going on here?
>

Ha. tleap definitely choked on your PDB file. You *do* in fact have just 1
GDP molecule, but it's split up over 2 different residues, separated by
water molecules. Residue 209 is GDP, and it has 28 atoms in it. Residue
214 is GDP, and it has 14 atoms in it. GDP itself is about 42 atoms large
(give or take a couple protons, depending on the protonation states of the
various phosphate groups). This is causing your issue. If you visualize
*only* residue 209, I would expect part of your GDP molecule to disappear (I
think ONLY your hydrogen atoms will disappear). Your GDP hydrogens appear
to be residue 214, whereas the heavy atoms appear to be residue 209. This
will cause headaches. Because now, in order to generate a psuedo-residue
fragment for GDP, PMEMD must create a fragment that spans residues 209
through 214, which spans multiple molecules; hence your error.

You can verify this yourself by looking at the RESIDUE_POINTERS flag of your
prmtop file. This basically tells you which atom number each residue begins
on. Thus, you can find the number of atoms in a residue by subtracting the
starting position of that residue from the *next* residue (and indeed, you
can look at ATOM_NAMES to see which atoms these are, exactly).

Can you maybe attach your PDB file that caused all of this grief so we can
get a better idea of leap's limitations?

All the best,
Jason


>
> Cheers,
>
> Maura
> ________________________________________
> From: Jason Swails [jason.swails.gmail.com]
> Sent: 17 June 2011 02:29
> To: AMBER Mailing List
> Subject: Re: [AMBER] PMEMD does not support intermolecular PRFs
>
> On Thu, Jun 16, 2011 at 8:54 AM, Maura Catherine Mooney <
> mmooney05.qub.ac.uk
> > wrote:
>
> > Hi Jason,
> >
> > Without a doubt, I agree with you. However, I wonder...this topology
> file
> > runs with sander (haven't checked the result - but it runs). I have
> checked
> > previous results with this and I don't see anything strange in the
> > trajectory. Would you expect to see 2 GDP molecules in the pmemd
> > trajectory, or is this just the way pmemd splits the workload. Also,
> > running with sander should be ok...right? (not desirable - esp because of
> > the speed) Am I right in thinking that sander does not implement
> > intermolecular PRFs?
> >
>
> All that happened was that there are 2 residues *named* GDP. That doesn't
> necessarily mean they are both GDP. You can use a visualization program
> (like VMD) to visualize your system, and highlight all resname GDP. This
> would show you what's happening. While sander runs without error (since
> pmemd and sander arrive at the same answer along 2 completely different
> paths). I would take the pmemd error as a warning that you need to fix
> your
> issues with your topology file.
>
> sander does not use PRFs, that's correct. sander splits the workload much
> more naively; splitting on actual residues rather than doing any kind of
> spatial decomposition with dynamic load balancing like pmemd does (which is
> a large part of why it scales so much worse)
>
>
> > Also, is there any easy way to learn how to read prmtops...is this
> detailed
> > in the manual? I guess a lot of it comes with experience...
> >
>
> You can study http://ambermd.org/formats.html carefully, but as far as I
> know it's not detailed in the manual anywhere. That'll give you a *basic*
> idea of what each part of the topology file means (but take care to read
> the
> italicized comments!).
>
> Understanding how to actually modify the topology file to get it to do
> somewhat more complex things is far more difficult, and requires more time
> invested in learning the format, as well as careful study of the codes that
> use the topology file. For instance, removing just one water molecule
> requires you to adjust just about every section of the topology file (you
> have to eliminate the atoms (atom_names, amber_atom_type, atom_type_index,
> etc.), adjust the residue_label, residue_pointer, nres, natom, charges,
> masses, atoms_per_molecule, solvent_pointers, etc.), and if you mess just
> one part up, your prmtop (if you're lucky), will crash pmemd and/or sander
> or (if you're unlucky) will run, giving you garbage results without you
> knowing.
>
> Even changing the charge of a single atom in the topology file requires you
> to remember that each atomic charge is multiplied by 18.223 before being
> written to the topology file. As is probably apparent, the topology file
> was *not* designed to be human-readable or easily modified.
>
> HTH,
> Jason
>
>
> > Cheers,
> >
> > Maura.
> > ---------------------------------------------------------------
> > Maura Mooney
> > School of Chemistry and Chemical Engineering
> > David Keir Building
> > Queens University Belfast
> > Stranmillis Road
> > Belfast
> > BT9 5AG
> >
> > mmooney05.qub.ac.uk
> > mauramooney9.gmail.com
> > ________________________________________
> > From: Jason Swails [jason.swails.gmail.com]
> > Sent: 16 June 2011 15:44
> > To: AMBER Mailing List
> > Subject: Re: [AMBER] PMEMD does not support intermolecular PRFs
> >
> > On Thu, Jun 16, 2011 at 8:39 AM, Maura Catherine Mooney <
> > mmooney05.qub.ac.uk
> > > wrote:
> >
> > > Hi Prof Case...and Jason,
> > >
> > > Many Thanx for the speedy reply.
> > >
> > > >[In general, careful editing of the input pdb file is required for all
> > but
> > > the
> > > >simplest protein or nucleic acid systems, and a careful review of what
> > you
> > > are
> > > >loading is always desirable.]
> > >
> > > Indeed so, I agree!
> > >
> > > >Then load your force field and any libraries needed for the GDP or Mg
> > > >moieties. Issue a single loadpdb command to create the basic system,
> > > >follow this by bond, solvateBox and addIons, etc. as appropriate.
> Then
> > > >examine the prmtop file that is created to make sure you have the
> > residues
> > > >and ions you want.
> > >
> > > This is what I have *intended* to do, but obviously, not *exactly* what
> I
> > > did. :) The tutorials are very good for the beginner, where a standard
> > > protocol is identified (which is what you mention above). This is what
> I
> > > did, although there were a system-specific tweaks, as always.
> > Additionally,
> > > I use xleap to prep the inpcrd/prmtop files and always use the 'check'
> > > command before creating these files - in your opinion, would you expect
> > > xleap 'check' to pick up such errors?
> > >
> > > FYI, I created this prmtop, using the method you specify above (and
> that
> > > which is detailed in the tutorial). The only difference here is that I
> > > implement PM3 qm for the GDP/Mg. ESP charges are calculated in
> gaussian
> > and
> > > then the RESP program is used in AMBER. The pdb file is then edited to
> > > contain the calculated charges, with a few minor changes (to make the
> > system
> > > neutral) and resulting lib file is saved. The fact that the prmtop
> > *thinks*
> > > there are two GDP in the residue labels but only one in the
> > > ATOMS_PER_MOLECULE confuses me. Can anything RE the source of the
> > problem
> > > be taken from this specific info, or is it simply...the topology file
> is
> > > wrong and needs fixed?
> > >
> > > Finally, the fact that residue labels seems to think there are 2 GDP
> > > molecules - is this likely the source of the PMEMD error...I don't know
> > > anything about intermolecular PRFs, but think I read that this could be
> > due
> > > to overlapping atoms. Is this true?
> > >
> >
> > The RESIDUE_LABELs aren't used for anything outside of specifying amber
> > mask
> > selections and *maybe* some other GB variable initialization. The real
> > problem is ATOMS_PER_MOLECULE. When splitting the workload, pmemd is
> > assigning 2 different molecules to the same pseudo-residue fragment. In
> > particular, RESIDUE_POINTER array shows that the *second* GDP residue is
> 19
> > atoms large, or so, yet ATOMS_PER_MOLECULE will show you that this
> residue
> > is split into several molecules. *This* is probably what I would say is
> > causing issues.
> >
> > Again, your best bet is to get rid of anything weird in your PDB file and
> > recreate your topology.
> >
> > HTH,
> > Jason
> >
> > Many Thanx for the advice.
> > >
> > > Maura
> > > ---------------------------------------------------------------
> > > Maura Mooney
> > > School of Chemistry and Chemical Engineering
> > > David Keir Building
> > > Queens University Belfast
> > > Stranmillis Road
> > > Belfast
> > > BT9 5AG
> > >
> > > mmooney05.qub.ac.uk
> > > mauramooney9.gmail.com
> > > ________________________________________
> > > From: David A Case [case.biomaps.rutgers.edu]
> > > Sent: 16 June 2011 15:13
> > > To: AMBER Mailing List
> > > Subject: Re: [AMBER] PMEMD does not support intermolecular PRFs
> > >
> > > On Thu, Jun 16, 2011, Maura Catherine Mooney wrote:
> > > >
> > > > I have little/no experience reading topology files, but I had a look,
> > > > based on your response. Strange behaviour indeed, as there is only
> > > > *intended* to be one GDP molecule, one Mg and 4 coordinating H2O's.
> > >
> > > You need to review how you created the topology file in the first
> place.
> > > I recommend creating a single PDB file with exactly what you want in
> > > it. Change any H2O or HOH labels to WAT; put a TER card between
> separate
> > > molecules. Remove CONECT records (but be sure to issue a "bond"
> command
> > > later to leap for any disulfide bonds or other non-standard bonds.) It
> is
> > > best to put all the water molecules at the end. If there are alternate
> > > conformations in file, choose the one(s) you want and remove the rest.
> > >
> > > [In general, careful editing of the input pdb file is required for all
> > but
> > > the
> > > simplest protein or nucleic acid systems, and a careful review of what
> > you
> > > are
> > > loading is always desirable.]
> > >
> > > Then load your force field and any libraries needed for the GDP or Mg
> > > moieties. Issue a single loadpdb command to create the basic system,
> > > follow this by bond, solvateBox and addIons, etc. as appropriate. Then
> > > examine the prmtop file that is created to make sure you have the
> > residues
> > > and ions you want.
> > >
> > > ...good luck.....dac
> > >
> > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> > > _______________________________________________
> > > AMBER mailing list
> > > AMBER.ambermd.org
> > > http://lists.ambermd.org/mailman/listinfo/amber
> > >
> >
> >
> >
> > --
> > Jason M. Swails
> > Quantum Theory Project,
> > University of Florida
> > Ph.D. Candidate
> > 352-392-4032
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Candidate
> 352-392-4032
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jun 17 2011 - 00:00:04 PDT
Custom Search