[gmx-developers] Re: [gmx-users] Oh topology .. go away ...

[Berk Hess]

>From: Marc Baaden <baaden at smplinux.de>
>To: "Berk Hess" <gmx3 at hotmail.com>
>Subject: Re: [gmx-users] Oh topology .. go away ...
>Date: Thu, 04 Apr 2002 14:38:02 +0100
>
>
> >>> "Berk Hess" said:
>  >> I think it is already possible to add an [atomtypes] section
>  >> between
>  >> #include ff...
>  >> and
>  >> [moleculetype]
>  >> in your topology file.
>
>Sorry Berk, but right now the confusion is on the rise again :)
>So what you say is:
>
>- I can add a second (3rd, 4th .. ?) [atomtypes] section after
>   eg ffgmx, but it has to be BEFORE the FIRST [moleculetype]
>   section ?
>
>- So, let me really be dumb and ask explicitly, can there be
>   several versions of the following sections and what are the
>   restrictions on how to place them (assumption: before the
>   first occurence of moleculetype):
>
>[ defaults ] ?
>eg can you switch generation of pairs/use c6,c12 and sigma,epsilon
>alternatively for everything that follows until the next occurrence
>of defaults ..

Well, grompp will warn you when you add a second defaults, it says
it will use the last one. But I just realized this might give unwanted
result when gen-pairs is set to yes. I'll change this warning into a
fatal_error.

All the [ ...types ] directives can be used as many times as you like.
But the atomtypes that are used in the other types should be defined
before the interaction types are defined.
I just found out that you can get serious problems with the LJ
parameters. There should only be one [ nonbond_params ] section in
your topology and it should be after all [ atomtypes ] directives,
as grompp generates the LJ parameters for all atomtype combinations
each time it encounters a [ nonbond_params ] directive. If no
[ nonbond_params ] directive has been found LJ parameters will be
generated at the first [ moleculetype ] directive.

Order in the bondeds in unimportant, both in the types section
and in the moleculetype section.

>Indeed I tried the atomtype thing and used a separate file with atomtypes
>after ffgmx. I get the same energies, but the topology files are
>different. gmxcheck says
>[..]
>
>do you know where that comes from and why that is ?
>is it a problem ?

This is probably due to the overwrites of explicit combination
by the generation with the combination rule.
So it probably is a problem.

>
>Meaning that one might evtl. also need a "user_nonbondeds" ..
>or additional nonbond_params entries

Indeed, additional nonbond_params entries.
The only problem that needs to be fixed in grompp is that the
LJ parameter generation should not overwrite the parameters
given in [ nonbond_params ] sections. I'll see if I can fix this.

>
>Sorry to keep on asking these questions, but I feel I have to
>understand it to be confident that I set up the correct topology
>for my molecule .. (and even in general, I think some errors have
>been committed in the past in our group, and this might help me
>spot them)

Don't be sorry. This things are very unclear, not described in the
manual, and probably unknown even to all the authors of Gromacs.
A serious problem is that the current grompp does not give warnings
or errors for invalid input (although I have already added a huge
amount of check on the input since Gromacs 2.0). But I think we're
almost there.

Berk.


_________________________________________________________________
Send and receive Hotmail on your mobile device: http://mobile.msn.com