[gmx-users] Re: stepsize too small ... but potential energy negative!

[Justin A. Lemkul]

Anna Marabotti wrote:
> Dear Justin, dear Luca,
> here's the answer to your questions:
> - I'm currently using the "classical" forcefield gromos96 43a1 (choice "0"
> in pdb2gmx). After producing the topology, the only warning I see from
> pdb2gmx is this one: WARNING: there were 0 atoms with zero occupancy and 63
> atoms with occupancy unequal to one (out of 1583 atoms). Check your pdb
> file.
> However, atoms with occupancy <1 are present also in "regular" PDB files (if
> I remember well, also in PDB files I used previously). Is this a problem?
> -  Box: I'm currently using a cubic box and I'm setting 1 nm of distance
> between solute and box with option -d (and also center the box). Looking at
> the system I can't see any contact between the protein and the box walls. I
> started by setting 0.8 nm and the problem was the same.
> - .mdp file: here it is:
> cpp                 =
> define              =  -DFLEXIBLE
> constraints         =  none
> integrator          =  steep
> nsteps              =  50000
> emtol               =  1000
> emstep              =  0.1

This step size is far too large!  Try something more like 0.01.

> nstlist             =  5
> ns_type             =  grid
> rlist               =  1.0
> coulombtype         =  PME
> rcoulomb            =  1.0
> rvdw                =  1.2
> fourierspacing      =  0.12
> fourier_nx          =  0
> fourier_ny          =  0
> fourier_nz          =  0
> pme_order           =  4
> ewald_rtol          =  1e-5
> optimize_fft        =  yes
> Tcoupl              =  no
> Pcoupl              =  no
> gen_vel             =  no
> It's very similar to the one Justin suggested in its tutorial.

...except for numerous changes.

> - PR-MD: I'not interested in skipping the minimization and continue with MD.
> I tried to launch a PR-MD step to see if the error produced in this step was
> more informative, and try to understand what was the problem on my
> structure. I failed, however...
> 
> I would also add that this is not the first time that I'm using an homology
> model produced by MODELLER to perform MD. All the checks I made on my model
> tell me that it is a good model, so I really don't understand what's wrong
> with it.
> 
> Finally, I started from the structure deprived of the first residue in order
> to see if that residue was the "bad" one, but the problem still persists and
> the potential energy has the same negative value with the same order of
> magnitude.
> 

Likely due to a flawed .mdp file.  The algorithm is trying to take too large of 
a step along the potential energy surface, causing bad geometry and infinite 
forces.  A bit more finesse should solve this problem.

-Justin

> Anna
> 
> 

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

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

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