AMBER Archive (2006)

Subject: Re: AMBER: Failure of iwrap in PMEMD

From: Robert Duke (rduke_at_email.unc.edu)
Date: Fri Jul 21 2006 - 08:34:02 CDT

Hi Myunggi -
Okay, I did a 0.5 nsec simulation overnight on 4 pentiums, amber 8, and
basically confirmed what you are saying. I looked at the wrapping code. It
is identical to the amber 6 sander wrapping code. In amber 7 the sander
code was changed without explanation; I thought it was a performance issue
(and I have not been that worried about wrapping performance, so I did not
track the change). So there may be a hole in the wrapping algorithm. An
important point here is that this in no way affects the validity of results
from pmemd - wrapping is only a visulization aid and is only done to the
mdcrd trajectory, not to any internal data. So this simply means that
currently to get a clean trajectory for visualization, you need to go
through ptraj (which apparently most people do anyway instead of using
iwrap - it is more efficient to wrap on one processor rather than having a
roomful of cpu's waiting for the master to do the wrapping, and wrapping is
not a parallelized process at the moment). I will check this out further
and issue fixes for amber 8 and 9 (I will probably just end up adopting the
new code). So there is no need to send your trajectory. And there is no
need for anyone to worry about their results. Thanks for pointing this out,
though.
Best Regards - Bob

----- Original Message -----
From: "Myunggi Yi" <myunggi_at_gmail.com>
To: <amber_at_scripps.edu>
Sent: Thursday, July 20, 2006 8:24 PM
Subject: Re: AMBER: Failure of iwrap in PMEMD

> Dear Dr. Duke,
>
> As you see, my system is solvated DMPC bilayers with a drug molecule.
> I did exactly same simulation with sander, and I have some experience
> also.
> I'm sorry for not sending you my trajectory.
> The file is big. I couldn't attach it.
>
> If you run with PMEMD with the restart file and input (iwrap=1) file
> (I did with IBM SP4 with 8cpu's),
> you will see some water molecules out of the unit cell (about after
> 500ps).
> Compare with the sander (iwrap=1) result.
> I know I see some part of the lipid chain outside the box, but not
> whole lesidue.
> However, I saw several water molecules (one lesidue) outside the box.
> I have never seen like this before using sander.
>
> I will try to send my trajectory to you later.
>
> Thank you and have a nice day.
>
>
> On 7/20/06, Robert Duke <rduke_at_email.unc.edu> wrote:
>> Myunggi -
>> Okay, I have looked at this. Your system has got several thousand waters
>> with several thousand residues of some sort, looks like roughly 13C
>> aliphatic, with several oxygens, all individual molecules, in a band in
>> the
>> middle of the simulation cell. In a sense there is no solute; just a
>> couple
>> of different solvents with significantly different polar character. I
>> noted
>> 1 other small molecule in there somewhere. The restart file looks "in
>> the
>> box" to me looking at vmd, and it is a 50 psec snapshot I think. Now,
>> for a
>> molecule to be wrapped by the wrapping routine, it's geometric center has
>> to
>> go outside the box. This means for the aliphatic stuff, you will see
>> half
>> chains waving around outside the box. I don't know where the actual box
>> boundaries are, but generally the edges look more dilute because you are
>> not
>> looking through as much solvent. So I did two 100 psec runs with pmemd
>> 8,
>> one with iwrap off, and one with iwrap on. I took 50 trajectory
>> snapshots
>> over the 100 psec. What I see when iwrap is off is that the water does
>> diffuse out, but the hydrocarbon stuff does not; the visual effect is of
>> a
>> dogbone where the ends are growing, if you know what I mean. With iwrap
>> on,
>> what I see is all the molecules retained in the simulation cell, though
>> the
>> long carbon chains are more visible for the molecules that are less than
>> half "out of the box". Now, if it takes a longer simulation, fine, but
>> why
>> don't you send me the mdcrd you have rather than me doing a 0.5 nsec run
>> just to reproduce this. Looking at the code, I don't see how there would
>> be
>> a problem, but this simulation is not the usual case, so maybe something
>> else strange is going on. When you say that you have not seen problems
>> with
>> sander, have you run this exact system with sander? My bet is sander
>> would
>> do the same thing, since the code is pretty much the same, but I don't
>> yet
>> see a problem with pmemd.
>> Regards - Bob Duke
>>
>> ----- Original Message -----
>> From: "Myunggi Yi" <myunggi_at_gmail.com>
>> To: <amber_at_scripps.edu>
>> Sent: Thursday, July 20, 2006 11:04 AM
>> Subject: Re: AMBER: Failure of iwrap in PMEMD
>>
>>
>> > Thank you all,
>> >
>> > Dr. Duke,
>> >
>> > I have never seen this before using sander.
>> > This system is not big, and as you see this is only the equilbration
>> > step.
>> > I saw this less than 500 ps short simulation.
>> > For the big system size and long time simulation, I have to use iwrap.
>> > As you know, the coordinate larger than 999.999 can't be formatted out,
>> > and the program crashs.
>> >
>> > I will send you my trajectory, topparm, and input/output files.
>> > By displaying with VMD, you will see the waters outside the primary
>> > unit cell (orthogonal).
>> >
>> > Have a nice day.
>> >
>> >
>> > On 7/20/06, Robert Duke <rduke_at_email.unc.edu> wrote:
>> >> Myunggi -
>> >> I am not aware of any problem with the code, and see no problem with
>> >> your
>> >> input. What type of unit cell are we dealing with here (orthogonal
>> >> vs.
>> >> truncated octahedron)? Anything else wierd at all? The pmemd
>> >> wrapping
>> >> code
>> >> is pretty much the same as the sander wrapping code I believe, but
>> >> will
>> >> have
>> >> to check. Are you sure that you would not see the same thing if you
>> >> ran
>> >> for
>> >> as long in sander? If you send me inputs/outputs (a small number of
>> >> trajectory frames for output, of course) I will take a further look.
>> >> I
>> >> believe a lot of folks do wrapping primarily with ptraj these days. I
>> >> have
>> >> seen examples of strange simulations where even the restrt file format
>> >> was
>> >> overflowed by parts of the simulation being strongly repelled, but
>> >> when
>> >> that
>> >> happens you obviously have something wrong going on. Is this a system
>> >> that
>> >> has been simulated for a really really long time by any chance?
>> >> Regards - Bob Duke
>> >>
>> >> ----- Original Message -----
>> >> From: "Myunggi Yi" <myunggi_at_gmail.com>
>> >> To: <amber_at_scripps.edu>
>> >> Sent: Thursday, July 20, 2006 10:01 AM
>> >> Subject: AMBER: Failure of iwrap in PMEMD
>> >>
>> >>
>> >> > Dear Amber users,
>> >> >
>> >> > I found the failure of image control (iwrap) in PMEMD (amber8).
>> >> > As time goes by, more and more water molecules came out of
>> >> > the primary box in my simulation.
>> >> >
>> >> > I have never seen this before using SANDER (iwrap).
>> >> > The following is my input, and I haven't found any error,
>> >> > warning, and weired thing in the output file.
>> >> >
>> >> > ==========================
>> >> > npt eq
>> >> > &cntrl
>> >> > nstlim = 1000000, dt=.002,
>> >> > irest=1, ntpr=500, ntwx=500, ntx=5,
>> >> > temp0=310.0, ntt=1,
>> >> > tautp=2.0, taup=2.0, cut=9.0,
>> >> > ntb=2, ntp=2, iwrap=1,
>> >> > ntc=2, ntf=2,
>> >> > /
>> >> > ==========================
>> >> >
>> >> > I know this won't effect the simulation, and
>> >> > I can handle this using ptraj in the post analysis.
>> >> >
>> >> > Is there anything wrong in my input?
>> >> >
>> >> >
>> >> > --
>> >> > Best wishes,
>> >> >
>> >> > MYUNGGI YI
>> >> > ==================================
>> >> > KLB 419
>> >> > Institute of Molecular Biophysics
>> >> > Florida State University
>> >> > Tallahassee, FL 32306
>> >> >
>> >> > Office: (850) 645-1334
>> >> > http://www.scs.fsu.edu/~myunggi
>> >> > -----------------------------------------------------------------------
>> >> > The AMBER Mail Reflector
>> >> > To post, send mail to amber_at_scripps.edu
>> >> > To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>> >> >
>> >>
>> >>
>> >> -----------------------------------------------------------------------
>> >> The AMBER Mail Reflector
>> >> To post, send mail to amber_at_scripps.edu
>> >> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>> >>
>> >
>> >
>> > --
>> > Best wishes,
>> >
>> > MYUNGGI YI
>> > ==================================
>> > KLB 419
>> > Institute of Molecular Biophysics
>> > Florida State University
>> > Tallahassee, FL 32306
>> >
>> > Office: (850) 645-1334
>> > http://www.scs.fsu.edu/~myunggi
>> > -----------------------------------------------------------------------
>> > The AMBER Mail Reflector
>> > To post, send mail to amber_at_scripps.edu
>> > To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>> >
>>
>>
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber_at_scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>>
>
>
> --
> Best wishes,
>
> MYUNGGI YI
> ==================================
> KLB 419
> Institute of Molecular Biophysics
> Florida State University
> Tallahassee, FL 32306
>
> Office: (850) 645-1334
> http://www.scs.fsu.edu/~myunggi
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu