[gmx-users] Computing work from pulling simulations

[Lutz Maibaum]

On Jun 21, 2017, at 2:32 AM, Leandro Bortot <leandro.obt at gmail.com> wrote:
> 
> Have you compared the results of your calculation using the forces at
> different intervals? e.g. every 1, 10, 50, 100 steps ?
> 
> As long as you use a not-so-large interval the result should be the same
> within statistical error.

Yes, that is of course a possibility. I get about 2% relative error between writing out the forces at every timestep vs. writing them our only every 1000 timesteps. That’s not the end of the world, but I was hoping to avoid it if Gromacs would keep track of the cumulative work performed by the umbrella potential.

Thanks for your reply!

Best,

Lutz



> On Mon, Jun 19, 2017 at 7:19 PM, Lutz Maibaum <lutz.maibaum at gmail.com>
> wrote:
> 
>> I am using “umbrella pulling” to simulate the dissociation of two
>> molecules. I would like to calculate the work W(t) that the harmonic trap
>> has performed on my system up to timestep t, which can be calculated as
>> 
>> W(t) = sum_{i=0}^t F_i * v * dt
>> 
>> where F_i is the biasing force at timestep i, v is the pulling velocity,
>> and dt is the timestep. The latter two I set in the .mdp file, and the
>> force is written out by Gromacs to the  pull_pullf.xvg file.
>> 
>> This works very well. However, computing the work using the equation above
>> requires me to save the forces to a file _every single timestep_, which
>> becomes unfeasible for longer simulations. Does Gromacs keep track of the
>> cumulative work performed by the umbrella potential, and does it save that
>> value somewhere? I couldn’t find it in the .edr or log files (I am using
>> Gromacs 5.1.4).
>> 
>> Best,
>> 
>> Lutz