 |
Tips for Users |
The following is a list
of tips that have been found to be useful to users during past experience.
Topology
Builder Section
-
Always check that the PDB file
you wish to use contains a fully connected molecule; for complexes, split
the file in several bits, each fully connected, and run DLGEN accordingly;
-
Avoid using gromos force field
with Ewald electrostatics, rather use Coulomb electrostatics with cut-off.
Molecular
Dynamics Section
-
Check the list of shortest interatomic
distances in the OUTPUT file, to avoid "crazy collisions" at initial step;
-
In cases of SHAKE/QUENCH failures
or for sudden crashes, always try to reduce the timestep; a short run (10
steps) with much smaller time step (even 10^-8 ps) will allow
proper adjustment of constraints; during the initial phase, always observe
potential energy; If potential energy is too high, perform a sequence of
short runs with increasing time step, until reaching a comfortable value
of potential energy; often "zero structure optimization" works at best;
-
Avoid using the CONTROL option
"ewald precision" for large systems, it would result in a CPU killer; for
similar reasons, prefer the SPME over Ewald option for large systems;
-
Investigate the k_MAX and alpha
dependence of electrostatic energy to choose optiomal values;
similarly for the number of grid points for SPME;
-
Avoid Ewald/SPME methods with
non-neutral systems;
-
Prefer "shift coulomb" electrostatics
rather than crude cut-off;
-
Use octahedric boundary conditions
for hydrated globular proteins;
-
Use the "fast solvent" option
to reduce CPU cost;
-
Don't alter the box dimensions
by hand for molecular systems, this would result in failure of the "utailor"
(i.e. reconstruction of molecules) algorithm!
-
Check conservation of energy
as dE/dK<10^-2 (E: total energy, K: kinetic energy) and no significant
drifts of energy over long times;
-
Check conservation of total
momentum after using the "cap" option, and generally before production
runs;
Analysis
Section