[gmx-users] Recommended way to run production using the same conditions as equilibration

[Justin A. Lemkul]

Andrew DeYoung wrote:
> Hi,
> 
> In molecular dynamics I have learned that there are three main phases:
> energy minimization, equilibration (say, 2 ns in duration), and
> production/dynamics (say, 3 ns in duration).  Suppose that I want my
> production run to use the same conditions as equilibration.  What is the
> best way to do this?
> 
> I can think of two conceptual ways to approach this:
> (1) Do one long 5 ns run.  Then, when doing analysis, ignore the frames
> corresponding to the initial 2 ns, using the -b option found in most of the
> Gromacs trajectory analysis programs to consider only the final 3 ns.
> 
> (2) Do equilibration and production in separate runs.  This way, doing
> analysis is perhaps more foolproof:  I don't need to worry about always
> neglecting the initial 2 ns, and the initial time I consider is always,
> conveniently, t = 0 ps.  
> 

If the conditions for equilibration and production are identical, then either 
approach works.  Most workflows separate equilibration and production due to the 
use of restraints, different T or P coupling schemes, etc.  But if all you're 
doing is extending the same ensemble, then the choice here between 1 and 2 makes 
no difference.

>>From looking at various Gromacs tutorials, it seems that approach (2) is
> recommended.  But now, even within approach (2), there are at least two
> options.  One option is to feed the checkpoint file from equilibration into
> the grompp part of production.  For example, 
> 
> EQUILIBRATION: 
> [equilibration has been prepared using grompp]
> mdrun -deffnm npt-eq -cpo npt-eq.cpt
> PRODUCTION:
> grompp -f npt-md.mdp -c npt-eq.gro -t npt-eq.cpt -p water.top -o npt-md.tpr
> -po npt-mdout.mdp
> mdrun -deffnm npt-md
> 
> where I have used the -cpo option in mdrun of equilibration and fed this
> resulting checkpoint file to grompp of production, using -t.  Is this
> correct?
> 

Yes.  See, for instance:

http://www.gromacs.org/Documentation/How-tos/Extending_Simulations#Changing_.mdp_file_options.

and

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

> Or, I could eliminate the grompp step of production (since my production run
> uses the same parameters as in equilibration) and feed the checkpoint file
> directly to mdrun of production:
> 
> EQUILIBRATION: 
> mdrun -deffnm npt-eq -cpo npt-eq.cpt
> PRODUCTION:
> mdrun -deffnm npt-md -cpi npt-eq.cpt
> 
> where I have used -cpi in mdrun of production to feed in the checkpoint
> file.  (In contrast to grompp, there is no -t option in mdrun.)  Is this
> correct?  My concern here is that, in the entry for mdrun in the Gromacs
> manual, it says that for the option -cpi, "By default the output will be
> appending to the existing output files."  But since I have specified output
> file names using -deffnm npt-md, will the trajectory indeed be written to
> npt-md files, or will it be appended to the npt-eq files, from which the
> checkpoint file originated?  
> 

They will be written to npt-md.xtc/log/etc.  When supplying -deffnm, the use of 
any other flag for input or output overrides the name in -deffnm, so you can use 
the previous checkpoint file, even if it has a different name.  You can also get 
around the appending problem by using -noappend.

> If both of these options are correct, which is preferable?
> 

Whichever one makes sense to you.  I tend to go with the grompp/mdrun combo. 
That offers you the ability to change .mdp options for I/O purposes, remove 
restraints, etc.

-Justin

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

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

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