AMBER Archive (2005)

Subject: Re: AMBER: SHAKE error

From: Agata Jurkiewicz (aggresja_at_gmail.com)
Date: Fri Jul 22 2005 - 17:06:49 CDT


Dear Dave,

The phosphates are the whole story. I always have some problems with
them since all the ligands I am working with contain phosphate groups
and I don't want extra hydrogens on them.
This molecule was prepared using Gaussian (DFT geometry optimization,
then HF/6-31G*, MK charges) and then Antechamber (parm99 forcefield).
I made some modifications to the force field (applying analogy to the
angles and torsion angles where they were absent), but as far as I
remember, these modifications were not necessary in the case of dUMP.
There are no protons connected to the phosphate groups. Sometimes I
need to add the protons during the geometry optimization of the
ligands, but I cleave them before calculating charges. In the case of
dUMP, I was able to perform geometry optimization without the protons
and the final geometry looked OK to me. It works for the simulation
with the protein.
I've had problems with phosphate groups using earlier versions of
Gaussian - I was not able to calculate charges for the anions and I
had to cleave hydrogens after, which resulted in non-integral charges
in the MD simulation. However, with Gaussian 03 it seems to work
better...
But maybe the initial geometry od dUMP is a problem which I can't see
at once, examining the pdb file? Should I run vacuum calculation with
dielc=4 to test the initial geometry?

Returning to the MD problem: I've just tried slower heating (in 4
steps, 25 ps and 75 K each) and nscm=50 during equlibration. The
equlibration failed on SHAKE even faster than before.
There was one simulation (taup=0.1, tautp=0.1, nscm=500) which did not
produce the SHAKE error but required sander restart. This restart also
failed:

 NSTEP = 1450 TIME(PS) = 54.950 TEMP(K) = 299.81 PRESS = -284.8
 Etot = -5613.5470 EKtot = 1303.2600 EPtot = -6916.8070
 BOND = 2.1875 ANGLE = 18.7547 DIHED = 17.4731
 1-4 NB = 8.8459 1-4 EEL = -106.3205 VDWAALS = 967.4099
 EELEC = -7825.1576 EHBOND = 0.0000 RESTRAINT = 0.0000
 EKCMT = 648.1449 VIRIAL = 815.2850 VOLUME = 27182.9359
                                                    Density = 0.8067
 Ewald error estimate: 0.2046E-03
 ------------------------------------------------------------------------------

 vlimit exceeded for step 1495 ; vmax = 131.934782789446
 vlimit exceeded for step 1498 ; vmax = 1.296717054530640E+019
 SANDER BOMB in subroutine nonbond_list
  volume of ucell too big, too many subcells
  list grid memory needs to be reallocated, restart sander

I suppose I can't go on restarting the equlibration every 3 ps or so.
I am going to try the same simulation on another of my ligands (all
pretty similar to dUMP, though). Maybe there is some problem just with
this one.

Regards,
Agata.

On 7/22/05, David A. Case <case_at_scripps.edu> wrote:
> On Fri, Jul 22, 2005, Agata Jurkiewicz wrote:
> >
> > Thank you for the hints.
>
> I'm wondering now if there is not some problem with the force field. You
> said originally that the molecule was dUMP. How did you make this? Are
> there any protons on the phosphate?
>
> I'm just making a guess here: we had problems with terminal phophates, where
> the geometry could get quite distorted (leading to shake errors) with
> "standard parameters". So, you might look closely at the phosphate
> geometries.
>
> If this guess is correct, you shold see problems even with a vacuum simulation
> with no waters. Of course, I am stretching here, but it would be good to know
> if there is anything non-standard about your molecule.
>
> ...dac
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber_at_scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu