Re: Sybyl FF followon
Date: Fri, 26 Jun 92 16:47:38 EDT
Sender: chemistry-request-: at :-ccl.net
Errors-To: owner-chemistry-: at :-ccl.net
In response to Joe Leonard's questions...
1) The cutoff is not required for single-molecule or non-
bounded calculations; the energy is the most accurate if the
cutoff is set to something larger than the molecule.
Of course you pay the penalty in time required since the
number of non-bonded pairs increases.
For simulations with periodic boundary conditions some kind
of cutoff is necessary, or Ewald summation is used in some
2) I agree with Damodaran in that 8 Angstroms might not be
suitable for all cases, but it seems to work for most
situations. The vdw energy drops off very rapidly and the
energy difference with and without cutoffs is small.
I'm an empiricist, so I did some calculations to see how
big a difference cutoffs make. Here are some energies computed
for crambin (1crn) with all hydrogens added, without any
minimization. (TRIPOS 5.2 force field, but Hydrogen vdw radius
is 1.2 Angstroms; Kollmann all-atom charges, constant
SYBYL, like AMBER uses a residue-based cutoff; if any atom in
a residue is within the cutoff distance from another, all
atoms in both residues are used in the non-bonded calculations.
Cutoff vdW E Electrostatic E # pairs
100 691.245 -2684.925 197,287
8 692.740 -2683.961 91,773
The 100 Angstrom cutoff takes about twice as long and increases
the accuracy by about 0.2%. Bernard Brooks did a lot of work
to see how different cutoffs affected dynamics trajectories.
I believe this is what Martin Norin is referring to.
3) There has been one change since the TRIPOS force field
paper, involving the handling of hydrogen bond pairs. In the
paper the vdw term for the two atoms forming the hydrogen
bond was set to zero so that the two atoms can move closer
together and lower the electrostatic energy. We found that
in dynamics the two atoms involved in the hydrogen bond had
a tendency to "fuse" so that the coulombic energy was
infinite; now we scale down the sum of the vdw radii used
in the vdw term by 70%.