[gmx-users] Periodic Cell of Cyclohexane--LINCS Warnings

[Justin Lemkul]

On 6/13/17 12:13 AM, Billy Williams-Noonan wrote:
> "Side issue, not related to your problem - use gmx grompp -t state.cpt to
> get the full-precision coordinates and box."
> 
> I used gmx energy to plot the box coordinates over time.  Using P-R with a
> 2 fs timestep and tau_p of 2 ps, the box shrinks to a certain point, and
> then the system "blows up".  This can also be visualised in VMD.  There is
> no such problem when increasing tau_p to 6 ps.  Thanks, I think I know what
> the problem is.
> 
> "At minimum, your use of rvdw = 1.0 is incorrect.  It should be 1.4.  This
> could be related to instability of systems dominated by vdW interactions."
> 
> Thank you, will set rvdw to 1.4 nm.  I initially assumed 1.0 nm would be
> fine, as that is how I read the mdp settings page for Gromacs*:*
> 
> rvdw (1) <http://manual.gromacs.org/online/mdp_opt.html#nl>
> 

Do not conflate a default setting required for a syntactically correct input 
file with a scientifically valid setting.

> When I found out that my rvdw was wrong, I noticed that many people seem to
> be using the wrong value for the cut-offs and still getting published.  I
> understand that the varying force-fields were optimised under different
> conditions.  My system appears to be stable now with the current
> parameters, but I will fix to make the long-term energetics more correct
> anyway.
> 

It's not a matter of "making things more accurate/correct" - the cutoffs are 
part of the force field, and if you get them wrong, it's likely that your 
simulation is invalid based on an imbalance of forces because you're violating 
the assumptions in the force field, most of which rely on a cancellation of 
errors.  And indeed, people publish wrong things all the time.  It's up to the 
researcher to do his/her homework about the force field, how it should be 
employed, and what it's caveats are, in designing the experiment.  Too often MD 
simulations are done via "cut and paste" from tutorials, unrelated work, etc. 
The literature is rife with junk, unfortunately.  Be skeptical!

-Justin

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

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

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