[gmx-users] Strategy for chaining jobs using differentmdp parameters (long)

[Berk Hess]

>>In order to chain the jobs, do i need to reference from the processed top 
>>and last gro from the previous step. In other words, do i need to use :
>>
>>pr2 step:
>>grompp \
>>     -p ../pr1/molecule_pr1.top \
>>     -c ../pr1/molecule_pr1.gro \
>>
>>instead of (as above):
>>pr2 step:
>>grompp \
>>     -p ../ini/molecule_ini.top \
>>     -c ../ini/molecule_ini.gro \
>
>I do use the new structure file in all of my chained jobs. The coordinates 
>in the structure file won't be being used because a .trr is supplied. For 
>an ensemble switch, the box size is not stored in the .trr IIRC, and the 
>only source of this information in the absence of an .edr file is the 
>structure file (hence the use of genbox). In your particular example, the 
>pre-NVT box will be the same size as the post-NVT box , so you shouldn't 
>have a problem here - but check it.

When you supply a trr file, the coordinates, velocities and box
will always be taken from the trr file.

>
>The output from -pp should be functionally identical to its input, by 
>construction, so you can use any .top here. I omit the use of -pp and just 
>reuse the same .top all the time. This stops you hacking something once by 
>hand and forgetting to change other occurrences.
>
>>I presume i need to use the previous step top and gro parameters, but 
>>using g_desort (not done yet).
>>Again in this case, is g_desort sufficient ?
>>Are -renum and -rmbdunbs only used for the run (but not permanently 
>>affecting the top and gro files)?
>
>Don't know.

grompp never modifies any of the input files.

Berk.

_________________________________________________________________
Live Search, for accurate results! http://www.live.nl