Hi Jake, I recently parameterised a small molecule for Amber as follows. I
ran Gaussian98 to fully optimise the structure to get the bond
lengths/angles etc, then ran RESP to get teh charges. For the dihedrals I
fixed one bond at a time at a specified dihedral - 0,15,30-180 etc, and
optimised the rest of teh structure. This gave me the true dihedral energy
plot for each angle. I then created the molecule in xleap, and input the
charges etc. I created a parameter file (like parm98) for my atom-types,
bonds, angles, dihedrals etc, but input the dihedrals as 0 force
constants. I then took a series of structures with the specified dihedral
at 0,15,30 etc and everything else identical, and calculated a one step
energy minimisation to give me the calculated energy of the molecule. This
gave me the non-bonding interactions as calculated by Amber, the next step
is to input the dihedral parameters that when added to the Amber energy
plot, give the true Gaussian energy plot. This info is on the Amber home
page under parameterisation. In my view the key is not neccessarily to use
the correct empirical values for the parameters, but values that make the
molecule behave as much like it should as possible. The forcefield can
never be perfect, so using fudge factors to make it better is justified in
my oppinion (perhaps I am opening myself up for a slating here). I hope
this is of help,
-----------------------------------------
Simon Cross
School of Chemistry
University of Nottingham
tel. 0115 9514193
Email: pcxsc_at_nottingham.ac.uk
Received on Wed Feb 14 2001 - 04:40:15 PST
