AMBER Archive (2007)

Subject: Re: AMBER: error in running sander

From: Thomas Cheatham (tec3_at_utah.edu)
Date: Tue May 29 2007 - 22:03:36 CDT


> Thanks a lot for your replies. Well, the same error msg appears even
> if I have deleted Mg-ATP from the pdb. As suggested, I looked at my
> initial pdb structure,
>
> ATOM 919 CG1 VAL A 126 15.020 10.602 2.155 1.00 40.87 C
> ATOM 920 CG2 VAL A 126 14.615 12.941 2.762 1.00 43.82 C
> ATOM 921 N GLY A 128 17.832 10.976 0.712 1.00 52.78 N
> ATOM 922 CA GLY A 128 18.873 10.542 -0.212 1.00 59.95 C
>
> However, I found that after 126 it is 128, but this sequence has been
> corrected in 'tleap' as:
>
> ATOM 1831 C VAL 126 36.583 33.672 43.611
> ATOM 1832 O VAL 126 37.744 33.723 44.012
> ATOM 1833 N GLY 127 36.305 33.201 42.383
> ATOM 1834 H GLY 127 35.343 33.143 42.082
>
> Is this the problem?

"Looking" at the PDB, i.e. the text file, will not likely be of much use.

The post from Ross Walker suggests that you *visualize* the structure with
molecular graphics (VMD, Chimera, SwissProtViewer, ...) and try to see if
atoms are overlapping.

You can also use the "checkoverlap" command of ptraj; assuming you have
starting coordinates named "inpcrd" and a prmtop file named "prmtop"

  ptraj prmtop
  trajin inpcrd
  checkoverlap

...this will print out a list of overlapping atoms (with distances < 0.95
A). See the manual for further information.

[The residue #'s you mention are not likely a problem; the algorithm in
LEaP looks for changes in residue #/name to decide when a new residue
starts; the exact sequence number is not important (or in other words,
when detecting the end of a residue, as long as the number is different
things are OK and a new residue is detected, the order does not matter)].
Also, LEaPs recognition of residues does not check things like atomic overlap or
look for errors in the coordinates.

Finally, if you have serious overlap problems, and still want to use the
LEaP produced inpcrd/prmtop (rather than making sure your input
coordinates are OK), you can try to first minimize with electrostatics
turned off; this can be done with the NMROPT=1 options and setting the &wt
for electrostatics (or even nonbonds) to be zero for the first round of
minimization. Beware that this too can fail, i.e. if a bonded strand
sits in the middle of a aromatic ring, the minimization will reduce the
energy, but you'll still have a strand stuck in the middle of the aromatic
ring since minimization cannot-- in proper usage-- cross barriers...

    \
    -\-\
    | \ |
     \---
        \

Short summary:

(0) make sure you can run and fully understand the tutorials. Jumping
into Fe-S clusters requires a fair bit of understanding of AMBER,
molecular mechanics, force fields, and simulation... Take heed the advice
of Walker/Simmerling to try the protein alone first.

(1) check your input coordinates carefully; visualize the AMBER built
structure with molecular graphics and also try out the checkoverlap
command of ptraj...

(2) errors usually result from problems with the input files, starting
coordinates, name mismatches, errors, ... Program error is possible, but
unlikely as these programs and their ancestors have been widely used for
at least a decade.

(3) if there are overlapping atoms, fix the input PDB file (usually there
are naming issues or multiple structures or ...) and rebuild the
inpcrd/prmtop.

(4) if you use tricks to relax the structure, such as shutting off
electrostatics, make sure you understand why there was a large energetic
penalty to begin with...

--tec3

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu