Re: [AMBER] TI in vacuum

From: FCUP <Sílvia>
Date: Tue, 8 Feb 2011 09:43:24 +0000

Hi Tom,

Does the same thing occur on different machines?

I put the simulation running in 2 different machines and the error still
occurred.

If you use just two CPUs without a queueing system?

These simulations are running already without queueing system.

If you slightly change lambda?

Well, here started again the "strange" results:

- where it was lambda=0.4, I change to 0.42, and where it was lambda=0.1, I
change to 0.12. In both cases the simulations went well until the end.

However, I decided to try a more slight change. So, where it was lambda=0.1,
I change to 0.0999. And it worked too... for purine->purine-0-ol but not for
phenanthrene->
2hydroxiphenanthrene (which had the same problem)... But, in the latter, the
simulation ran with lambda=0.099....

I also tryed to change lambda=0.4 to 0.400 and 0.40000 (in
Imidazole->Imidazole-5ol), but it didn't work in either.

Which strategy should I follow?

Thanks in advance,
Slvia Martins



2011/2/2 <steinbrt.rci.rutgers.edu>

> Hi,
>
> well the pdb look fine but doesnt help much.
>
> >>> > In the second transformation, it stops at lambda=0.1, and the v0
> >>> stops
> >>> > with
> >>> > no error message and v1 terminates normally, again.
>
> hm, one process dying and the other running to completion? That should not
> be possible, since each process requires info from the other to advance.
> This makes it sound like a problem of running sander as opposed to a
> problem with the chemical model. Does the same thing occur on different
> machines? If you use just two CPUs without a queueing system? If you
> slightly change lambda?
>
> >>> > I am not writing a trajectory file for each lambda. For each lambda,
> >>> i
> >>> am
> >>> > just writing a coordinate file that corresponds to the last
> >>> structure.
>
> Try and run one that you know will fail and write out the trajectories
> each step, then compare V0 and V1 for e.g. coordinate mismatches.
>
> Thomas
>
> Dr. Thomas Steinbrecher
> formerly at the
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>



-- 
Slvia Alexandra Pinto Martins
PhD student
Chemistry Department
Faculty of Sciences
University of Porto
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Feb 08 2011 - 02:00:02 PST
Custom Search