[gmx-users] Extending Simulations

[Justin Lemkul]

On 6/26/13 4:12 PM, HANNIBAL LECTER wrote:
> It is not very clear as to what you are trying to do? If you have the final
> coordinates and velocities from a previous simulation, I would recommend
> start a simulation using the .gro file and ensuring that you have
> gen-vel=no in the .mdp file.
>
> If you have a well-equilibrated system, then the properties should not vary
> much. On similar lines, you can stitch multiple .edr files using eneconv.
>

Coordinates and velocities, especially in single precision, are insufficient to 
preserve the previous simulation ensemble.  The .cpt file contains a lot more 
information, especially related to continuity of thermostats, barostats, etc.

>
> On Wed, Jun 26, 2013 at 3:54 PM, Neha <nshafique at wesleyan.edu> wrote:
>
>> Hi everybody,
>>
>> I have a question that relates to reproducibility in Gromacs. If I have a
>> .cpt file from a previous simulation and use that for two simulations using
>> tpbconv, how similar should the temperature graphs etc be? When I use the
>> .edr files to plot various properties, there is significant difference
>> between the two graphs and I am not sure how much is acceptable or
>> expected.
>>

Have you seen both of these articles?

http://www.gromacs.org/Documentation/Terminology/Reproducibility
http://www.gromacs.org/Documentation/How-tos/Extending_Simulations#Exact_vs_binary_identical_continuation

The initial states should be identical, but the trajectories can quickly 
diverge, as is the case with any MD simulation.  The observed properties, on 
average, should converge to comparable values.

>> Along the same lines, is there any way to get exactly (or as close as
>> possible) same trajectories etc that I am missing out on? I am passing the
>> .cpt file to mdrun but are there other niggling details I should take into
>> consideration?
>>

mdrun -reprod can help in this regard.  Performance will likely take a hit, 
though.  What's not clear to me is why you *need* identical trajectories.  If 
you want to produce two identical trajectories, what's the point of running two 
at all?

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================