RE: AMBER: Problem with minimisation of an protein-inhibitor complex

From: Ross Walker <ross.rosswalker.co.uk>
Date: Tue, 23 Nov 2004 10:16:04 -0800

Dear Daniel

> VDWAALS = 2168.2854 EEL = ************* HBOND =
> 0.0000

Your structure is very bad as the electrostatics are off the chart. Normally
such problems would come about because you have atoms far too close together
but strangely your van der waals energy seems reasonable. Thus you may want
to check the charges by hand - perhaps you have an atom with an abnormally
large charge (misplaced decimal point?).

>
> To find where the structure is bad I tried to fix groups of atoms
> by using iBelly. So I used the following configurations :

This is not going to work as your forces are so huge. Plus fixing the system
will prevent you from finding a minimum. Anyway, when you do find a minimum
it will only be with those atoms fixed. As soon as you start MD the system
will fly apart. Besides the above error in the electrostatics suggests there
is something fundamentally wrong with your structure, minimisation is not
going to be able to fix it.

> So there were less problems calculating the energy but I did not get a
> minimized structure.

There were not problems before with calculating the energy. The EEL is
probably very very big. The *'s are printed because the value is too big to
fit in the space allocated for it in the output file. Internally (assuming
you haven't overflowed a REAL) it is stored properly and used.

> So I thought it could help to perform an MD because I assumed
> the complex
> could stick an a local minimum and got the following :

There is no way your system would stick in a local minimum with MD with such
a huge EEL.

> vlimit exceeded for step 40; vmax = 1.32912849E+14
> vlimit exceeded for step 41; vmax = 1.10446189E+13
> vlimit exceeded for step 42; vmax = 7.04650286E+13
> vlimit exceeded for step 43; vmax = 2.13182016E+12
> Frac coord min, max: 0.00643121772 1.00398667
> The system has extended beyond
> the extent of the virtual box.
> Restarting sander will recalculate
> a new virtual box with 30 Angstroms
> extra on each side, if there is a
> restart file for this configuration.
> EWALD BOMB in subroutine Routine: map_coords (ew_force.f)
> Atom out of bounds. If a restart has been written,
> restarting should resolve the error

Your system has blown up and flown off into space... What you could try
simply as a guide to tell you where to look at your structure is to set up
MD for say 500 steps with a step size of 0.01fs (dt=0.0001)! Set ntwx=1 and
then it will dump te coordinates on every step. You should then be able to
load the mdcrd file into something like VMD and watch where your system
starts to blow up from. Then you will be able to see what part of the system
is duff...

> My questeions to the community are :
>
> Is it a good method to fix groups of atoms to find "ill" regions of my
> complex ?

Probably not. It is far quicker and more efficient to try a few 100 steps of
MD with a very short time step and write the coordinates on every step. The
human eye is very good at spotting abnormalities in visual data.

> What would you do in my case ?
> Do I have a big bug within my configuration files ?

I would start by verifying that your amber installation is good - check the
test cases pass. Then:

I would take a careful look at the prmtop file first of all - check what the
charges sum to. See if any atom has a strange charge. If this all looks good
I would then take a look at the initial structure - use ambpdb to get a pdb
from the prmtop and inpcrd file. Look at it carefully and systematically.
Try simplifying the structure - e.g no ligand and see if you can get a
stable simulation.

All the best
Ross

/\
\/
|\oss Walker

| Department of Molecular Biology TPC15 |
| The Scripps Research Institute |
| Tel:- +1 858 784 8889 | EMail:- ross.rosswalker.co.uk |
| http://www.rosswalker.co.uk/ | PGP Key available on request |

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Tue Nov 23 2004 - 18:53:01 PST
Custom Search