> the production run (10000 structures). I used the following
> input script
> to minimize each frame:
> maxcyc = 20500, ncyc = 1000,
> 1. The minimization is going very very slow - on an altix
> machie, using 4
> processors. In 48 clock hours, after submitting the job, only
> ~4600 frames
> (out of 10000) have been minimized. There are no other jobs
> running on the
> machine. Could there be any problem with the system so that the
> minimization is going so slow?
You don't say how long your 10ns simulation took but at 2fs this 10ns
simulation represents 5,000,000 MD steps. Then in your minimisation you are
asking for 20,500 steps which if every frame of your 10,000 structures took
the full 20,500 steps then you are asking for a total of 205,000,000 so you
can expect your complete set of minimisations to take around 41 times longer
than your 10ns MD simulation took. Is this roughly what you are seeing? If
so then nothing is wrong.
Although one might question why you want to minize the 10,000 structures and
if you are looking to locate specific minima on the energy surface is a gap
between structure samples of only 1ps appropriate? Also is necessary for you
to run such a huge number of minimisation steps? Do you really need to get
that close to the minima in each case?
> 2. I would like to get the coordinates of the minimzed
> structures. How do
> I specify this in the sander.MPI command? B'se with reference to the
> previous discussions on AMBER reflector, -x is the flag for
> reading in the
> trajectory file. I also checked the trajene files of the test
> set provided
> with AMBER. In this set also, I could not figure out where
> the minimized
> structrues are stored. In the minimiztion run that I am
> doing, even the
> .rst file has not been created.
The coordinates from a minimization get written to the restart file as
specified with the -r flag so in your case:
mpirun -np 4 $AMBERHOME/exe/sander.MPI -O -i cga_gb_fnl_min.in -p
../../prmtop -o cga_gb_fnl_min.out -c ../cga_gb_prod10ns.rst -x
../cga_gb_prod10ns_trj.crd -r cga_gb_10ns_fmin.rst > out_cga_gb_fnl_min
cga_gb_10ns_fmin.rst which if you use the same name for all 10,000
minimisations just gets overwritten on every run such that you only have the
coordinates for the last calculation that ran. it is probably blank because
the run is still in progress when you looked at it. (-x is not required in
minimization)
Note also that by specifying -c ../cga_gb_prod10ns.rst as the input
structure for the minimisation, assuming you used the same rst structure for
each case you will have successfully run the same minimisation 10,000
times... There is no function to read the coordinate file in minimisation.
You will need to use ptraj to split your mdcrd file into 10,000 sequentially
numbered rst files. You will then need to script the running of your
minimisation such that each run picks the next sequentially numbered input
rst file and writes a unique (probably sequential) -r rst file on each run.
You can then post process all these rst files produced from the minimisation
in ptraj or ambpdb to obtain pdb files if you need them.
All the best
Ross
/\
\/
|\oss Walker
| HPC Consultant and Staff Scientist |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
|
http://www.rosswalker.co.uk | PGP Key available on request |
Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Jun 21 2006 - 06:07:10 PDT
