Re: CCL:More AMBER atom types

[longshot "at@at" chem.duke.edu (Brad Isbister)]

 Charles Letner writes about the selection of atom types in AMBER.
 The selection of atom types should always use a data set which resembles the
 compounds to be investigated.  The original AMBER parameters which were
 derived for proteins and nucleic acids are pretty good for the functional
 groups found in such systems (peptide backbone, AA sidechains, C, T, G, A, etc).
 If your system includes other functional groups, a literature search for AMBER
 parameters which better reflect the 'real' system is a good idea.  These new
 atom types can be included in an frcmod file for addition to the standard
 parameter set.  Several papers have been written on 'non-standard' functional
 groups.  In my work, I use the parameters for carbohydrates from S.W Homans,
 Biochemistry, 1990, 29, 9110-9118.
 If you want to know just what compounds were used to derive the standard AMBER
 parameter set, get the original AMBER papers:  Kollman, Case, et al, JACS,
 1984, 106, 765-784 and Kollman, Case, et al, J. Comp. Chem., 1986, 7, 230-252.
 You should also look in the AMBER manual p23-24 for more info on the parameter
 databases.
 I've seen rumblings of "Does anyone(else) want to set up a database of
 published AMBER parameters (or references to such)?"  But I don't think
 anyone has taken the plunge and volunteered.
 There is always the possibility that your compound is very different from
 anything you can find.  In that case, refer to some papers on deriving atom,
 bond, angle, and torsion parameters from spectroscopic data and/or Quantum
 Mechanical simulation.  I haven't had to go that route yet.  Perhaps some
 others on the CCL can help out with the references?
 In the end, Molecular Mechanics is a game of assumptions.  Too many guesses and
 your results are worthless or misleading; too few and you'll be running very
 long and inefficient simulations.  You'll have to determine how similar the
 database compounds are to the compound of interest.
 Good Luck!
 -Brad
 ---------------------------------Clip N Save------------------------------------
 Brad Isbister					Duke University
 longshot "at@at" chem.duke.edu				Department of Chemistry
 Computational/Biophysical chemistry		E.J. Toone group