[gmx-developers] Making personal plugin

[Jonathan Khao]

Anton Feenstra a écrit :
> Anton Feenstra wrote:
>> Mark Abraham wrote:
>>> Jonathan Khao wrote:
>>>> Hello !
>>>>
>>>> We're currently trying to develop a tool for gromacs which returns 
>>>> the system's coordinates in a cylindrical coordinate. We're newbies 
>>>> in the gromacs developement ... The trxio.h seemed to have the 
>>>> right fuctions to do so.  The problem is that it includes a lot  of 
>>>> dependencies and making tests  and debugging would take an eternity 
>>>> if  we'll have to recompile  the entire  thing.  Are there any 
>>>> instructions somewhere on the web on how to make fast tests while 
>>>> programming or any documentation on how to tweak gromacs ? If not , 
>>>> could you give us some hints or instructions ?
>>>
>>> You haven't explained well enough what you're trying to do... make 
>>> an analysis tool or not? If the former, then your testing cycle need 
>>> only rebuild the executable + objects + libraries that you're 
>>> actually changing, by "make <executable>" in the src/xxx subdirectory.
>>
>> If what you want is extract coordinates in a cylindrical (i.e., 
>> polar?) representation, I don't see that you need to modify trxio at 
>> all. It should be sufficient to write a small tool that reads 
>> trajectory, calculates cylindrical coordinates, and writes them out. 
>> Presumably, this could be done (re-)using one of the existing 
>> trajectory formats, since the dimensionality doesn't change. (But you 
>> should be careful not to read them as carthesian...) You should also 
>> be careful about precision issues, which are non-homogeneous in 
>> cylindrical/polar.
>>
>> What you *might* want to add to the gmxlib are two (simple) routines, 
>> one that converts carthesian coordinates to cylindrical, and another 
>> that does the opposite.
>
> Ah, by the way, in case you need it, some of the basic maths can be 
> found in pdbio, where (carthesian) gromacs box (9 matrix elements, or 
> three box edge vectors) is converted to crystal unitcell 
> representation as used in pdb (three angles and three edge lengths).
>
>
Thank you for your responses. Effectively , I haven't been clear on what 
we intend to do.  My system is a Protein Detergent Complex, and our goal 
is to know how atoms behave are along the faces of an alpha helix. So we 
wanted to make a tool that read a Trr file, and converted the 
coordinates  as a function of :
        - the position along the axis of the protein ( fitted to the Z 
axis )
        - the distance from the axis
        - and an angle
So correct if I'm wrong, as you suggested, the use of gromacs's 
functions to read the "trr" trajectory is not worth it, a custom made 
tool should therefore be more appropriate ? If so are there any 
instructions on how to "decrypt" the "trr"  files ?

I've also been working on an other project for some time. I wanted to 
get properties of some molecules (detergents) which are in the vicinity 
of my alpha helix. 
The properties that I want to get are avaliable using the gromacs tools. 
But where I need help, is in the selection of the molecules in the 
"vicinity".

I've done verry heavy programming scripts which computes the distances 
of each molecules ( degergents ) with the g_mindist , then i read each 
of the "xvg" files , got the intervals where it was under the cutoff 
value, made sub trajectories, concatenated them and then applyed the 
gromacs tools. I suppose there are easy shortcuts to follow , especially 
by avoiding the time spending printing and reading parts.
So could you help me by  telling me how i could apply a filter on the 
tools ? It would be verry usefull, and would at last help me to dive in 
the gromacs source code.

Thank you again

-- 
Ph.D. Jonathan KHAO
LISM-IBSM-CNRS
31 Chemin Joseph Aiguier
13402 Marseille Cedex 20
France
jkhao at ibsm.cnrs-mrs.fr
tel : 0616738391