They won't be the same as a single long run, ever. FPRE is amazing in what it can do over a million or more iterations...
That said, I guess they're all sampling conformation space so if you threw away the first indeterminate number of steps to minimize correlation between them as a result of sharing the same initial state, it'd be qualitatively equivalent. But I'm not expert enough to detect when that correlation is gone...
-----Original Message-----
From: mdkvia.yahoo.com [mailto:mdkvia.yahoo.com]
Sent: Thursday, September 16, 2010 13:06
To: AMBER Mailing List
Subject: Re: [AMBER] Sander/pmemd usage: refc question
Hi Jason,
Thank you for the reply. I apologize for the confusion with -c. I meant -ref.
So my question #2 is: Say I run a single job with -ref option and the reference coordinates are in file ref.rst. Now I divide this single job to 3 as I mentioned in the previous email. So, now for the -ref of the 3 jobs I should provide the same ref.rst if I want to have the same results as the single long job. Is this line of thinking correct?
Yes, I understand that there will be rounding off errors etc. But apart from those, the results must be comparable b/w the single long job and the 3 shorter jobs. If not I was just wondering how to run a long simulation without having to deal with unmanageably huge trajectory files.
Thank you,
Maal
--- On Thu, 9/16/10, Jason Swails <jason.swails.gmail.com> wrote:
> From: Jason Swails <jason.swails.gmail.com>
> Subject: Re: [AMBER] Sander/pmemd usage: refc question
> To: "AMBER Mailing List" <amber.ambermd.org>
> Date: Thursday, September 16, 2010, 3:23 PM
> Hello,
>
> On Thu, Sep 16, 2010 at 3:10 PM, <mdkvia.yahoo.com>
> wrote:
>
> > Dear Amber users,
> >
> > I'm new to Amber and I would like to divide a MD job
> into several subjobs.
> > So, I intend to invoke sander.MPI (alternatively
> pmemd) the following way:
> >
> > sander.MPI -O -i subjob.inp -o subjob_1.out -c
> subjob_0.rst -p testjob.top
> > -r subjob_1.rst -ref subjob_0.rst -x subjob_1.mdcrd
> -inf subjob_1.inf
> >
> > sander.MPI -O -i subjob.inp -o subjob_2.out -c
> subjob_1.rst -p testjob.top
> > -r subjob_2.rst -ref subjob_1.rst -x subjob_2.mdcrd
> -inf subjob_2.inf
> >
> > sander.MPI -O -i subjob.inp -o subjob_3.out -c
> subjob_2.rst -p testjob.top
> > -r subjob_3.rst -ref subjob_2.rst -x subjob_3.mdcrd
> -inf subjob_3.inf
> >
> > In the subjob.inp file I'll be having ntx=5 and
> irest=1. My questions are
> > as follows:
> >
> > 1. If the simulation has no restraints (ntr=0), then
> do I even need to
> > specify -c option?
> >
>
> -c specifies the input coordinate (inpcrd) file, not the
> restraint file.
> The restraint file is specified by -ref . Therefore,
> -c is ALWAYS required
> (unless, of course, your input coordinate file is called
> "inpcrd"). -ref is
> only required when you specify ntr=1.
>
>
> >
> > 2. If I do (either because the answer to #1 is yes or
> ntr=1) then shouldn't
> > the coordinates for the -c option be same (say
> subjob_0.rst) for all 3
> > subjobs (i.e. above usage is wrong)?
> >
>
> Why would they have to be the same? It depends on
> what you're trying to
> do...
>
>
> >
> > 3. Lets say each subjob is 1 ns in length. Would the
> combination of the 3
> > trajectories be identical to a single 3 ns job (run
> with identical options
> > except for nstlim)?
> >
>
> What do you mean by "identical"? Round-off will make
> them slightly
> different for a long enough trajectory. However, it
> should behave more or
> less the same as a single long run. The numbers will
> not likely be
> "identical" except under strict circumstances. You
> can always check for
> yourself, though.
>
> Hope this helps,
> Jason
>
> Thank you in advance,
> >
> > Maal
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
>
>
>
> --
> Jason M. Swails
> Quantum Theory Project,
> University of Florida
> Ph.D. Graduate Student
> 352-392-4032
> _______________________________________________
> 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
-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information. Any unauthorized review, use, disclosure or distribution
is prohibited. If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 16 2010 - 13:30:05 PDT
