Re: splitting up long run leads to very different DNA conformation in E field

From: JC Gumbart (
Date: Fri Jul 14 2006 - 11:49:28 CDT

A couple things I notice...

Are you sure you sourced the right input files on this run? I notice
you have output name *c but inputname *a. Maybe you meant *b? Also,
while you didn't minimize here, could it be possible you minimized
again when you did your second run?

Finally, it's quite possible that it's just normal. Most of the
deviation comes from a dip in the distance early on; otherwise, the
lines are parallel. That indicates to me either a normal fluctuation
or a problem in restarting. You can check your output in VMD first
to see if anything looks unusual. And try running the shorter jobs
again, even with the same configuration files if you want, but break
it up differently, say like 600, 600, 600 ps.

On Jul 14, 2006, at 11:14 AM, Todd Trimble wrote:

> I haven't tried yet to reproduce the single vs. multiple run results.
> I'm running the single one again with smaller time step (1 fs). The
> configuration file used for one of the shorter runs is attached (Am I
> allowed to send attachments on this list?). I use the same
> configuration file for the single run, but with different restart
> files
> and higher number of steps. Also, I've attached a pdf file of the
> elongation results. The multiple runs were separated into run 1
> (0-50ps), run 2 (50-200ps) and run 3 (200-2000ps). Thanks for your
> help,
> Todd
> -----Original Message-----
> From: JC Gumbart []
> Sent: Thursday, July 13, 2006 6:56 PM
> To: Todd Trimble
> Cc:
> Subject: Re: namd-l: splitting up long run leads to very different DNA
> conformation in E field
> Is this problem reproducible to the same degree of difference? Could
> you give examples of your configuration files (the long run and, say,
> one of the restart runs)?
> Also, have you plotted the length over the course of the
> simulations? It would be helpful to determine the time course of the
> difference in length.
> On Jul 13, 2006, at 6:40 PM, Todd Trimble wrote:
>> Hi All,
>> I am trying to look at the effect of an applied electric field on
>> ssDNA
>> conformation (one end fixed). The solvent is treated explicitly,
>> along
>> with counterions that neutralize DNA charge (128 Na + 99 Cl ions).
>> The
>> ionic strength is high, as is the E field strength (10mV/ang). I'm
>> using the PME algorithm to handle the electrostatics. The problem
>> I'm
>> seeing is that I get very different results (i.e. end-to-end
>> distance)
>> when the simulation is carried out in a single run compared to when I
>> split it into several shorter runs. The end-to-end distances in
>> the 2
>> cases are different by about 50%. The shorter simulations use the
>> restart files (.coor, .vel, .xsc) of the previous simulation. The
>> temperature and pressure setpoints are 300K and 1atm. The system has
>> been through an extensive equilibration procedure prior to turning on
>> the field. The volume of the cell does not seem to change much
>> during
>> simulation, and the PME grid size is around 1 ang. or less. I am
>> also
>> using 2 fs/step with all H bonds set to rigid.
>> The initial electrostatic energy is identical to the final
>> electrostatic
>> energy from the previous simulation, but the initial potential
>> energy is
>> a little different than the final value of the previous
>> simulation. I
>> assumed that this was due to the regeneration of the PME grid at the
>> start. Because the volume changes so little it's hard to believe
>> that
>> the generation of a new grid periodically could alter my results so
>> drastically, but I can't think of anything else that is changing
>> when I
>> restart a simulation.
>> Has anyone else had similar problems? Any advice would be greatly
>> appreciated.
>> Todd
>> *********************************************************************
>> *
>> Dr. Todd M. Trimble
>> Asst. Professor of Research
>> Center for Applied Nanobioscience at the Biodesign Institute
>> Arizona State University
>> work: 480-727-0435
>> fax: 480-727-8283
>> *********************************************************************
>> *
> <ssDNA length.PDF>
> <30mer_H.conf>

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:42:21 CST