Re: Regular vs US sampling runs.

From: Vermaas, Joshua (Joshua.Vermaas_at_nrel.gov)
Date: Fri Sep 01 2017 - 11:53:11 CDT

You wouldn't happen to be using NAMD 2.11, would you? In that case, it would indeed be very slow. What I did for a similar kind of calculation was indeed to only use a selected number of atoms to compute the center of mass. However, since the COM calculation handles motion of lipids across the periodic image a bit less gracefully than might be desired, I would advocate for recentering your system so that the center of the bilayer is aligned to the center of the periodic cell before you start, and maybe using the atoms in the middle of your membrane to define the center.

As another word of warning: the z-position may not be the right reaction coordinate if you are interested in membrane binding. For larger molecules interacting with the membrane, the different potential orientations with respect to the membrane could get you radically different interactions, and as a result your convergence would be super slow (see 10.1021/jp501622d) as it needs to sample along an orthogonal degree of freedom that you aren't biasing. What I did when looking at membrane attachment was to use something like a contact number instead (coordnum in colvars, 10.1021/acs.biochem.6b00468), which has a much easier time converging.

-Josh

On 09/01/2017 10:32 AM, jeevan gc wrote:
Dr. Fiorin,

Thank you for your response.

There are 33 atoms in one group and 5888 atoms in another group.

I chose heavy atoms for the center of mass of the membrane. I am thinking of selecting few atoms from each lipid such as P, N, O etc. in order to reduce the number of atoms in group.

Thank you.

Jeevan

On Thu, Aug 31, 2017 at 7:34 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com<mailto:giacomo.fiorin_at_gmail.com>> wrote:
For 64 POPC molecules, you're talking about the center of mass of 3,300 atoms. Think about whether you can simplify that.

But in any case, centers of mass are computed in parallel (and used as such by Colvars starting in NAMD 2.12) and a few thousand atoms shouldn't be too much of a problem.

The Colvars output should tell you the size of each group of atoms. Can you report those numbers?

Giacomo

On Thu, Aug 31, 2017 at 9:13 PM, jeevan gc <gcjeevanbdr_at_gmail.com<mailto:gcjeevanbdr_at_gmail.com>> wrote:
Thank you Josh for quick and very informative response.

The system consists of a short helix and membrane (64 phospholipids). The reaction coordinate is defined as the center of mass between them. Both reference files are .pdb file with heavy atoms assigned occupancy of 1. I want to calculate the free energy of translocation across the bilayer.

Thank you.

Jeevan

On Thu, Aug 31, 2017 at 4:56 PM, Vermaas, Joshua <Joshua.Vermaas_at_nrel.gov<mailto:Joshua.Vermaas_at_nrel.gov>> wrote:
No. You probably have a colvar that is asking for waay too many atoms.
See
http://www.ks.uiuc.edu/Research/namd/2.12/ug/node56.html#SECTION000133400000000000000.Vermaas%40nrel.gov%7Cff909d95dfb64c29e11908d4f15709b9%7Ca0f29d7e28cd4f5484427885aee7c080%7C0%7C0%7C636398803510087946&sdata=wr62RHteTXWfKBLLmoekhUKjQHLplbP95y83h%2F9HJgw%3D&reserved=0>.
What does your colvar configuration file look like, and how many atoms
are being used in your reaction coordinate definition?

-Josh

On 08/31/2017 05:28 PM, jeevan gc wrote:
> <html>
> <head>
> <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
> </head>
> <body>
> <div dir="ltr">Dear NAMD Users,
> <div><br>
> </div>
> <div>I am running a Umbrella Sampling (US) Simulations using NAMD2.12 CUDA version &#43; a Quadro M5000 graphics card on 48 core linux machine. The all-atom membrane system has &nbsp;~32000 atoms.</div>
> <div><br>
> </div>
> <div>The regular run takes 6 hours to complete &nbsp;50ns. The same system in umbrella sampling simulation takes 35 hrs to complete 50ns. The longer US simulation time is due to additional harmonic restraint calculations.</div>
> <div><br>
> </div>
> <div>Is it normal for a US simulation time to be six times longer than a regular run?</div>
> <div><br>
> </div>
> <div>Thank you in advance.</div>
> <div><br>
> </div>
> <div>Jeevan</div>
> <div><br>
> </div>
> <div>PS: &nbsp;&#43;16 CPU processors, saved at every 20ps, colvar frequency 10,000<br clear="all">
> <div><br>
> </div>
> -- <br>
> <div class="gmail_signature" data-smartmail="gmail_signature">
> <div dir="ltr">
> <div>
> <div dir="ltr">
> <div>
> <div dir="ltr"><i><font face="times new roman, serif">Jeevan B. GC, Ph.D</font></i>
> <div><font face="times new roman, serif"><i>Post Doctoral Research Associate</i></font></div>
> <div><font face="times new roman, serif"><i>College of Pharmacy</i></font></div>
> <div><font face="times new roman, serif"><i>Washington State University</i></font></div>
> <div><font face="times new roman, serif"><i>Spokane, WA 99224 ,&nbsp;</i></font><i><font face="times new roman, serif">USA</font></i></div>
> </div>
> </div>
> </div>
> </div>
> </div>
> </div>
> </div>
> </div>
> </body>
> </html>
>

--
Jeevan B. GC, Ph.D
Post Doctoral Research Associate
College of Pharmacy
Washington State University
Spokane, WA 99224 , USA
--
Giacomo Fiorin
Associate Professor of Research, Temple University, Philadelphia, PA
Contractor, National Institutes of Health, Bethesda, MD
http://goo.gl/Q3TBQUgEGIhVjEqAkFJzLMxTdJ3JlxVtlA%3D&reserved=0>
https://github.com/giacomofiorin6&sdata=UInhjwhkeTQEKdQUEfTpLJnI4nIE0m%2F2MD6MDc2Mapg%3D&reserved=0>
--
Jeevan B. GC, Ph.D
Post Doctoral Research Associate
College of Pharmacy
Washington State University
Spokane, WA 99224 , USA

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:20:34 CST