Hi Peter,
Sorry if this comment is a little off target, as I have been on a long
vacation and not following this thread. But I wanted to note that I do
something that works okay for me to avoid this problem, though it is a
little tedious. I basically observe my system with vmd every 5 nsec or so
to start. When I observe that my complex has completely exited the primary
unit cell, I simply do a single step with iwrap on. This wraps the entire
complex back properly, and then I just observe the system at the shortest
time interval required to catch the primary unit cell exit. This can be a
bit trickier than it sounds, and not foolproof, because smaller molecules
can go out of range for the restart in the same time interval - if that is
happening, then you simply need to wrap more frequently, but only do the
wrapping when all complex molecules are in the primary unit cell. If you
just leave iwrap on all the time, then as individual molecules exit the
primary unit cell, they wrap individually, and this causes huge grief. I
guess one approach that would be cleaner than this would be to allow group
specification of a collection of atoms for wrapping in a future version of
amber, but that is a bit of a nuisance too. The existing binary restarts in
older versions of amber are badly nonportable, and I consider going back to
that format to be a non-starter. The use of netcdf would I presume work,
but I think it is good to have another option, as there can be issues with
installation and data inspection. We could introduce yet another fixed
format, with say at least 1000x larger range, and that would solve the
problem for the forseeable future, though the resulting files would not be
readable. I don't know if the vmd guys are already supporting reading
netcdf or not.
Best Regards - Bob Duke
----- Original Message -----
From: "Peter Varnai" <p.varnai.sussex.ac.uk>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Thursday, August 20, 2009 11:23 AM
Subject: Re: [AMBER] Distance restraint and periodic boundary conditions
> Dear Dave,
>
> Thanks for your response. It is hard to avoid the need to use nmropt with
> iwrap when one does umbrella sampling in explicit water for 100ns<. I
> think binary restart files (ntxo=0) are no longer allowed (amber10). It
> would be good to re-enact this option though. Until then I wonder if some
> tweak could force iwrap to consider the complexed molecules as one: pmemd
> has a variable "no_intermolecular_bonds" which could redefine the complex
> as one fused molecule if I introduced some covalent linkage (with zero
> force constant) between the subunits. To correct my restart files with
> separated subunits (to continue simulation) I will use ptraj and image the
> subunits back to a complex, write out this restart file and then paste the
> velocities from the old restart file... would be nice if ptraj could also
> dump the velocities from the restart file after imaging.
>
> best regards,
> Peter
>
> case wrote:
>> On Wed, Aug 19, 2009, Peter Varnai wrote:
>>
>>
>>> It is probably a good thread to share an odd experience using NMR
>>> distance restraints between two molecules and octahedral periodic box
>>> with iwrap=1.
>>>
>>
>> Thanks for the detailed report. The bottom line from my perspective is:
>> don't use iwrap=1 if you can possibly avoid it. In particular, don't use
>> iwrap=1 with NMR-like restraints in place.
>>
>>
>>> It is probably more complex that this as when a simulation starts from
>>> molecules in proximity and then iwrap separates them virtually during
>>> the simulation, it does not seem to affect the restraint energy in the
>>> actual run, but after restart the behaviour described above will happen.
>>
>> I think iwrap only affects output coordinates, not the internal ones the
>> program is using. This would explain the behavior you report.
>>
>>
>>> It became a bit tricky to run long simulations where I am forced to use
>>> iwrap but the imaging puts the molecules in a different arrangement back
>>> to the central box, eg. | A...B | can become | B...A|.
>>>
>>
>> For Amber developers: it seems to me that the "correct" solution here is
>> to allow binary restarts, so that no one is "forced to use iwrap". Plus,
>> we
>> could have wrapping code in just one place (ptraj), where there is logic
>> to
>> have greater control over how wrapping is done that we have in
>> sander/pmemd.
>>
>> So, is there something broken with binary restarts in the current
>> code? [The unformatted form doesn't have ltemp0; I don't know what else
>> might or might not be broken.] If someone could look into this (Scott?)
>> it would be a great help. This is important enough that we should
>> consider a
>> patch to Amber10, if needed.
>>
>> ...thanks...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
Received on Mon Aug 24 2009 - 11:11:11 PDT
