AMBER Archive (2008)

Subject: Re: AMBER: Questions about Locally enhanced sampling (LES) and Replica exchange

From: Carlos Simmerling (carlos.simmerling_at_gmail.com)
Date: Wed Aug 20 2008 - 12:47:12 CDT

see my comments below

On Wed, Aug 20, 2008 at 3:25 AM, xiaonan zhang <heptoking_at_gmail.com> wrote:
> Dear Amber users:
> After reading relevant papers and the AMBER 9 manual chapters on LES
> and REMD, I have a few
> questions on the implementation of these kind of MD.
>
> It seems to me tha LES method is relatively computationally cheaper than
> REMD, since the latter usually requires square(Number of atoms)
> independent replicas at different temperatures.

LES and REMD are often used for different purposes. REMD generates a
particular ensemble, where LES is often used to find the global
minimum.

> When the loop regions is
> bigger, that require quite a lot of replicas Is that right?

the # of REMD replicas usually depends on the size of the total sysm,
not a loop size. The # of LES copies doe snot need to depend on loop
size, just the desired # of alternate conformations that you want the
system to be able to sample at each step.

Another question
> pertaining to parallel computation of REMD is: if a REMD require N replicas,
> does this mean at least N cpu should be used simultaneously??

the # of replicas determines the # of MPI threads, which could be run
on different CPUs but can also be run on fewer CPUs. the exact details
depend on how your system is configured and the queueing software (if
used). In Amber, you set the number of "groups" with the ng flag, and
the # of MPI threads is usually set with a -np flag to mpirun. the #
of actual cpus is set by your MPI nodefile. If this isn't clear, then
you might check the MPI documentation or some of the examples.

> As Dr. Simmerling has noted in
>>
>> Generation of accurate protein loop conformations through low-barrier
>> molecular dynamics. Proteins. 2003 Jun 1;51(4) 577-90.
>
> A soft-core potential energy was used. However, I did not find documentation
> on this in the AMBER 9 manual. Does this option exist in this distribution?

unfortunately it does not, it was in an earlier version of amber but
never made it to the released code.
if you contact me offline I may be able to share the code if you have
a license to amber.

> As a non-expert in Molecular dynamics, I am pretty confused with the LES and
> REMD and various hybrid methods.

both of these methods are intended for people that are already quite
comfortable with standard MD, so you might want to work through the
tutorials etc before trying these.

> If my goal is to refine the loop structure
> of a homology model, which method is better to obtain a reasonable loop
> folding with a reasonable computation effort? It seems to me that LES is
> better since the convergent behavior of the copies will usually suggest a
> reasonable fold.

this can be true, especially if you are not interested in ensemble behavior.

> As there is no LES tutorial in the AMBER site and documentation is limited,
> It is a little bit difficult for me to grasp the LES method. Moreover, the
> Moil-view webpage: http://morita.chem.sunysb.edu/~carlos/moil-view.html is
> not reachable right now. More comprehensive documentation or guidance is
> needed. Are there any plans on this?

I will have to check on the web site status, we recently replaced the
server and it may not have all links working yet.
have you read the LES info in the amber manual? that and the test
cases and publications are all that we currently have available.
>
> Thank you for your attention.
> Xiaonan Zhang
> Shanghai public health clinical center
>
>
>
> 

-- 
===================================================================
Carlos L. Simmerling, Ph.D.
Associate Professor Phone: (631) 632-1336
Center for Structural Biology Fax: (631) 632-1555
CMM Bldg, Room G80
Stony Brook University E-mail: carlos.simmerling_at_gmail.com
Stony Brook, NY 11794-5115 Web: http://comp.chem.sunysb.edu
===================================================================
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
      to majordomo_at_scripps.edu