[gmx-developers] Amber force fields

[Erik Lindahl]

Hi,

I looked at the script Anton started writing, and there are a couple of 
things I spotted:

* The default parameters in Amber are NOT sigma/epsilon, but rather 
rmin/epsilon.
* Amber (and OPLS) uses a different version of the force constants in 
the force
   field equations. This applies to both bonds and angles.
* Some of the newer dihedrals have more than one term, so we might as 
well use R-B
   for everything.
* The dihedral definition isn't identical to gromacs for the normal 
2-periodic proper. It
   will differ by a factor of two if you just copy it.

In general, it's probably easier not writing very fancy scripts, since 
most things need to
be checked carefully manually anyhow. I hacked up an extremely simple 
perlscript that
translates the basic nonbonded and bonded interactions into Gromacs 
format - you
can find it attached in a tarball together with the Amber parm99.dat and 
the produced
output where I've added heavy hydrogen options and dummy constraints.


I won't put more work into Amber, but hopefully there are a couple of 
you willing to continue
the work - it shouldn't be too difficult. What remains to be done:

* Double- & Triple-check things to make sure everything is correct this 
far. There are
   a couple of atoms (Fe,Cu) where Amber99 doesn't provide L-J 
parameters, but you
   might be able to find them in earlier versions, or just copy the OPLS 
stuff.

   DO NOT try to understand the conversions from my script - derive them 
yourself so we
   are more confident if the two versions agree :-)

   I did not divide the dihedral coefficients by the number of dihedrals 
defined for a bond.
   There is a coefficient for this in the Amber force field, but I don't 
know if the recipe
   says all dihedrals should be used or not.  Check the literature.

* Write / edit the residue topology database. The only impropers you 
need are those defined
   in the OPLS rtp file, so just check that. For dihedrals I have no 
idea if the default Gromacs
   choices are fine, or if you need to define all dihedrals. Note that 
there is a new option to the
   rtp file (check ffoplsaa.rtp) so you can do this automatically 
without any pdb2gmx options...

* Fix hydrogen & termini databases. This really needs to be automatic if 
it should be part
   of Gromacs. Sorry, but I don't consider it an acceptable solution to 
edit each PDB file manually
   and change the residue names ....

Good luck,

Erik

-------------- next part --------------
A non-text attachment was scrubbed...
Name: amber99.tar.gz
Type: application/x-gzip
Size: 20348 bytes
Desc: not available
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20020615/66969ca8/attachment.gz>