[gmx-users] is there any reasonable way to produce a custom topology?

[Mark Abraham]

Yegor Isakov wrote:

You are much more likely to get a helpful response from the people who 
know this software best (the programmers, who do read this list) if you 
express a little less of your frustration, and imply less condemnation 
of their software design choices, which you probably don't understand 
yet anyway!

Check your assumptions - you haven't paid anybody for the right to rant 
at them :-) You'd also do well to remember that GROMACS doesn't come 
with a warranty of fitness for any purpose :-)

> I would be extremely pleased if anybody would tell me how could I foce 
> that stupid pdb2gmx program to follow the path to the directory 
> containing .itp files (~/share/top).

Setting the environment variable GMXLIB makes various GROMACS utilities 
look there instead of the default, which is probably 
$GMXDATA/gromacs/top. GMXDATA gets set when you run GMXRC which is 
installed in the same place as the GROMACS binaries. You may need to 
link the existing contents of $GMXDATA/gromacs/top/ to ~/share/top/ in 
order to have everything work, as I couldn't get GMXLIB to accept a path.

I note that the manual mentions GMXLIB only in passing in a few of the 
programs' man pages, and not at all in the Appendix section on 
Environment variables. Usage of GMXRC is also not mentioned in the 
installation section on the web page. Can either of both of these be 
looked at, please?

> Although in the manual it was recomended to avoid modifying .rtp files, 
> that is the only way I found to produce a topology "semi-automatically." 
> Is there any other  reasonable way to produce a topology? Is anywhere 
> more intelegent converter of pdb files? Is anywhere a program that 
> permits creating a stable bond between 2 atoms  in a topolgy without 
> blowing up the last one?

Construction of topologies for MM calculations is a process that 
involves making judgements. Automating "sensible" defaults is an 
extremely tricky business, given the multiplicity of force fields, the 
kinds of kludgey PDB files that you can get hit with, and the 
innumerability of the molecules people will want to use. Then of course, 
a default someone thinks is sensible, someone else will think is 
ludicrous, and they might both be right for different applications. 
Basically you will need to do some work yourself if you are doing 
something new... big surprise.

The limitation on one bonded interaction type per atom pair is not 
unreasonable. It probably stems from an implementation decision as much 
as a design choice. No force fields I'm aware of need this kind of 
thing. You can have a distance restraint and a bonded interaction, 
however, or different types of bonded interaction.

Mark