Re: AMBER: pmemd iwrap trouble

From: Robert Duke <rduke.email.unc.edu>
Date: Tue, 26 Aug 2008 00:38:40 -0400

Carlos -
I did have one idea on what may be going on here. For pmemd 10 (I presume
what you are using), in order to support extra points efficiently, I
introduced a "no_intermolecular_bonds" option. The default value for this
is 1, and what happens then is any molecules (as defined in the prmtop)
which have been covalently bonded are redefined as 1 fused molecule. Now, if
ptraj treats these differently in wrapping, there could be grief (ie., a
difference in how the wrapping occurs). One option, as long as you are not
using an extra points forcefield, is to set no_intermolecular_bonds to 0 in
&cntrl, and then the molecule definition will match the prmtop (but this may
not be permitted if you have tip4p water for instance). The use of
no_intermolecular_bonds was absolutely essential for high scaling with extra
points ff's, though you really could do tip4p and tip5p without this nicety
(where you would run into problems is with stuff in gaff I believe). I
actually also believe it to be the better interpretation of the situation,
but clearly this could create a ptraj/pmemd inconsistency. So the critical
question - would you have any molecules that would be fused in your system?
Regards - Bob

----- Original Message -----
From: "Carlos Simmerling" <carlos.simmerling.gmail.com>
To: <amber.scripps.edu>
Sent: Monday, August 25, 2008 12:35 PM
Subject: Re: AMBER: pmemd iwrap trouble


> Bob- a an aside to this, I'm having trouble with ptraj imaging pmemd
> traj files. I never had this with sander, but with pmemd many of the
> frames don't properly image a dimer. Do you know which ptraj image
> syntax/options match that used in pmemd? I'm using this, which always
> works for sander for an 830 residue protein (415 in each monomer). I
> want the dimer to be reconstructed properly.
>
> center :1-415 mass origin
> image origin center familiar
>
> but often the imaging isn't right- the rmsd for the dimer is a few A
> larger than it should be, and then it hops back to normal values. you
> can also see shifts in the dimer during the MD in VMD< then it hops
> back again. I know it's an imaging issue- do you know how to fix it?
> thanks
> carlos
>
>
>
> On Mon, Aug 25, 2008 at 11:45 AM, Robert Duke <rduke.email.unc.edu> wrote:
>> Hi Lars,
>> It should be perfectly compatible; I just checked over the code and iwrap
>> is
>> used regardless of box type for both restart and mdcrd. However, I do
>> believe that if ioutfm = 1 (ie., binary restart file), it is true that
>> wrapping does not occur. It is actually not needed in the case of a
>> binary
>> restart - overflow is not practically possible. We should probably
>> document
>> this "feature" or release a patch to change the behaviour (the problem is
>> caused by the fact that writes of the binary restarts occur on the actual
>> coordinates, whereas all wrapping is done on a temporary copy of the
>> coordinates because you don't want to actually wrap the coordinates
>> internally. This should not be affecting your trajectory, just the
>> restart
>> file. Since you see this with ioutfm 0, I suspect this is also an issue
>> of
>> the truncated octahedron not looking right in vmd, but am not at all
>> certain. Seems I once heard Darden say something about this stuff (I
>> think
>> he wrote the trunc. oct. wrapping code). Tom? I'll look at all the
>> relevant code in both pmemd and sander later in the week when I have my
>> source machine up, talk to other key amber guys, and post something to
>> the
>> list as to what, if anything we intend to do about binary restarts not
>> being
>> wrapped.
>> Best Regards - Bob
>>
>> ----- Original Message ----- From: "Lars SkjŠrven"
>> <lars.skjarven.biomed.uib.no>
>> To: <amber.scripps.edu>
>> Sent: Monday, August 25, 2008 10:58 AM
>> Subject: Re: AMBER: pmemd iwrap trouble
>>
>>
>>> Hi again,
>>> I did some more testing without any results.. It seems to me that
>>> iwrap = 1 is not compatible with octahedral water box?
>>> Lars
>>>
>>> On Mon, Aug 25, 2008 at 3:32 PM, Lars SkjŠrven
>>> <lars.skjarven.biomed.uib.no> wrote:
>>>>
>>>> Hi Bob,
>>>> Thanks for the quick reply. removing ioutfm=1 does not yield a
>>>> different result. nor changing to sander. :-/
>>>>
>>>> getting rid of some of my input variables the input-file looks like
>>>> this
>>>> now:
>>>>
>>>> Wrap
>>>> &cntrl
>>>> imin= 0, irest= 1, ntx = 5,
>>>> ntb = 1,
>>>> cut = 10,
>>>> ntc = 2, ntf = 2,
>>>> ntt = 0,
>>>> nstlim = 1, dt = 0.001,
>>>> iwrap = 1, ioutfm=0
>>>> /
>>>>
>>>> Am I sure it is not just an imaging problem?
>>>>
>>>> I think its not: In the multimer simulation (which gets a few subunits
>>>> displaced during the run with iwrap=1) the rmsd jumps from 3┼ to 80┼
>>>> when doing rmsd analysis in ptraj. for the monomer simulation the rmsd
>>>> does not yield an immediate jump, but continuing the simulation after
>>>> iwrap=1 yields and increasing rmsd value after a few more ns. maybe
>>>> implying that the protein has been translated and starts interacting
>>>> with an image?
>>>>
>>>> I will continue working on this and let you know if I find anything
>>>> else..
>>>>
>>>> Cheers,
>>>> Lars
>>>>
>>>>
>>>> On Mon, Aug 25, 2008 at 2:40 PM, Robert Duke <rduke.email.unc.edu>
>>>> wrote:
>>>>>
>>>>> Hi Lars,
>>>>> I am on vacation today, but will look at it tomorrow. There should
>>>>> not
>>>>> be a
>>>>> problem with wrapping in amber 10 pmemd; there was a bug in amber 9
>>>>> pmemd
>>>>> for which a patch was released. That said, I am not sure there is not
>>>>> some
>>>>> intricacy when binary trajectory files are in use, and I will have to
>>>>> look.
>>>>> It should not matter, as the restart file is the primary issue here,
>>>>> and
>>>>> it
>>>>> is not binary, but maybe there is some combination of inputs that
>>>>> screws
>>>>> up
>>>>> on wrapping. Are you sure it is not just an imaging problem? I have
>>>>> grief
>>>>> going back and forth between what pmemd or sander does and what ptraj
>>>>> does
>>>>> (but I am talking constant pressure here, and ptraj I think hits grief
>>>>> with
>>>>> the changing boxsize). Anyway, I will look into it, but you might try
>>>>> a
>>>>> 1
>>>>> step, no binary output run just for grins, or do a single step in
>>>>> sander
>>>>> and
>>>>> see if you get the same result (just gives me more info, in case
>>>>> something
>>>>> obvious does not jump out).
>>>>> Thanks - Bob Duke
>>>>>
>>>>> ----- Original Message ----- From: "Lars SkjŠrven"
>>>>> <lars.skjarven.biomed.uib.no>
>>>>> To: <amber.scripps.edu>
>>>>> Sent: Monday, August 25, 2008 6:25 AM
>>>>> Subject: AMBER: pmemd iwrap trouble
>>>>>
>>>>>
>>>>>> Dear Amber users,
>>>>>>
>>>>>> I've been running a MD-simulation for about 40ns and needed to wrap
>>>>>> the water back into the primary box (due to water has swimed too far
>>>>>> away). As recommended on the mailinglist I used iwrap for only one
>>>>>> step:
>>>>>>
>>>>>> Wrap it
>>>>>> &cntrl
>>>>>> imin= 0, irest= 1, ntx = 5,
>>>>>> ntb = 1,
>>>>>> cut = 10,
>>>>>> ntc = 2, ntf = 2, tol = 0.000001,
>>>>>> ntt = 0,
>>>>>> nstlim = 1, dt = 0.001,
>>>>>> ntpr = 1, ntwx = 1, ntwr = 5,
>>>>>> ioutfm = 1, iwrap = 1,
>>>>>> &ewald
>>>>>> dsum_tol = 0.000001,
>>>>>> /
>>>>>>
>>>>>> The restart file produced by this run has about 50% of the solute
>>>>>> outside the waterbox (when visualized in vmd). I can still carry on
>>>>>> my
>>>>>> MD-runs (with iwrap=0) after this, but I am worried about the new
>>>>>> coordinates for the solute with respect the the water.
>>>>>>
>>>>>> However, when I do reimage in ptraj it seems to be ok:
>>>>>> center :1-224
>>>>>> image familiar
>>>>>> trajout test.rst restart
>>>>>>
>>>>>> It must be said I use an octahedral water box. Both pmemd9 and 10
>>>>>> yields the same results for me with this respect.
>>>>>>
>>>>>> To wrap it up:
>>>>>> - should iwrap=1 with pmemd10 work when using a octahedral box ?
>>>>>> - or should I use ptraj to reimage? if so, are the velocities ok ?
>>>>>>
>>>>>> I realise this is a topic that has been discussed on the mailinglist
>>>>>> earlier and I apologise if my questions are redundant. I searched,
>>>>>> but
>>>>>> did not find a clear answer on this..
>>>>>>
>>>>>> Best regards,
>>>>>> Lars Skjaerven
>>>>>> University of Bergen, Norway
>>>>>> -----------------------------------------------------------------------
>>>>>> The AMBER Mail Reflector
>>>>>> To post, send mail to amber.scripps.edu
>>>>>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>>>>>> to majordomo.scripps.edu
>>>>>>
>>>>>
>>>>> -----------------------------------------------------------------------
>>>>> The AMBER Mail Reflector
>>>>> To post, send mail to amber.scripps.edu
>>>>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>>>>> to majordomo.scripps.edu
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> mvh Lars SkjŠrven
>>>> Institutt for Biomedisin
>>>> Universitetet i Bergen
>>>>
>>>
>>>
>>>
>>> --
>>> mvh Lars SkjŠrven
>>> Institutt for Biomedisin
>>> Universitetet i Bergen
>>> -----------------------------------------------------------------------
>>> The AMBER Mail Reflector
>>> To post, send mail to amber.scripps.edu
>>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>>> to majordomo.scripps.edu
>>>
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber.scripps.edu
>> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
>> to majordomo.scripps.edu
>>
>
>
>
> --
> ===================================================================
> Carlos L. Simmerling, Ph.D.
> Associate Professor Phone: (631) 632-1336
> Center for Structural Biology Fax: (631) 632-1555
> CMM Bldg, Room G80
> Stony Brook University E-mail: carlos.simmerling.gmail.com
> Stony Brook, NY 11794-5115 Web: http://comp.chem.sunysb.edu
> ===================================================================
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
> to majordomo.scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo.scripps.edu
Received on Wed Aug 27 2008 - 06:07:40 PDT
Custom Search