[gmx-users] box size

[Erik Lindahl]

Hi,
>
> I just built an ATP topology by just adding ATP to my ffoplsaa.rtp 
> file.
> I just listed the atoms and the bonding. How well does gromacs do at
> picking the correct angle and dihedral terms if they are not specified.
> More Specifically do the parameters gromacs chooses by default make any
> sense and how do I see what they are?
>
pdb2gmx always generates angles & dihedrals, but you can override it if 
you want.

Angles are essentially bullet-proof since every single force field I 
know of define them in the same way, and they always use all angles.

Dihedrals are a bit special in that some forcefields only use one 
dihedral per bond (Encad, Gromos96), but others generate every single 
possible combination - up to 9 dihedrals per bond (OPLS). There is a 
flag to control this in the header of the RTP file.

Improper dihedrals/Out-of-plane torsions are the real pain since every 
single force field seem to define their own standard for atom orders, 
so the safe solution is to list them explicitly. For united-atom force 
fields you use them both for planar groups and chiral centers, while 
many (but not all) united-atom force fields only use them for planar 
groups.


That's the first step - you can check the generated topology if you 
don't trust the program :-)

In the second step, grompp assigns parameters from the force field. 
This is a very simple rule-based process, and you will get a warning if 
the program cannot find the parameters for an interaction (it will say 
something about using zero for the parameters).

OPLS is again special in that some dihedrals have exceptions from the 
standard value when the atoms are located in certain residues, so in 
this case I have listed them explicitly in the rtp file. The non-listed 
ones will have the default values.


Cheers,

Erik