Dear Franz,
This makes sense, thank you very much for the help.
Kind regards,Miro
On Thursday, 21 July 2022, 11:13:00 BST, Franz Waibl via AMBER <amber.ambermd.org> wrote:
Dear Miro,
if your cpptraj was compiled with GPU support, the GPU implementation
will be used automatically unless you specify the "pme" flag (the
combination PME+GPU is sadly not implemented).
Best regards,
Franz Waibl
Am 21.07.2022 um 12:03 schrieb Miroslav Dikov via AMBER:
> Dear Franz,
> Thank you for the help, I think I am finally starting to get some sensible values for the average number of water molecules! I realise that 1000 frames is not enough, but I was just testing the functionality. I will give 100000 a try!
>
> Another quick question: I know that one can use the "pme" flag to run GIST with PME, but I can't see anything on a CUDA flag. Does it pick up an available GPU automatically?
> Many thanks.
> Kind regards,Miro
>
> On Wednesday, 20 July 2022, 13:27:46 BST, Franz Waibl via AMBER <amber.ambermd.org> wrote:
>
> Dear Miro,
>
> If you create your groups in the same way as shown in the tutorial, the
> input dx file is filtered, i.e., it contains a 1 for voxels that should
> be in a group, and and 0 for voxels that should not be in a group.
> Actually, the data needs to be filtered somehow, since gistpp searches
> for adjacent non-zero voxels.
>
> This translates to the group output. The grcount.txt file contains the
> number of voxels assigned to each group, while the group1.dx file
> contains the voxels of group 1 as a binary mask (as in the input file).
>
> You can use this mask to obtain the number of water molecules in a group
> (averaged over the frames, but summed over the group) like this:
>
> gistpp -i gist-gO.dx -i2 group1.dx -op mult -o gist-gO-group1.dx
> gistpp -i gist-gO-group1.dx -op sum | tee sum-group1.txt
> awk '/sum of:/{print "Number of waters in group 1:", $5 * 0.0329 *
> 0.125}' sum-group1.txt
>
> (Replace 0.0329 for your reference density, and 0.125 for the voxel volume.)
>
> You could also multiply your filtered input with the density before
> running group, to obtain the scaled groups directly. Just make sure that
> the low-density regions are still set to zero.
>
>
> Unless this is a testing run, I would recommend using more than 1000
> frames in the GIST analysis. Optimally around 100000, but we often use
> 10000 for large systems. Use the PME or GPU implementations to speed up
> the calculation.
>
> Let me know if this helps.
>
>
> Best regards,
>
> Franz Waibl
>
> Am 20.07.2022 um 12:52 schrieb Miroslav Dikov via AMBER:
>> Dear AMBER users,
>> Recently I started using GIST as implemented in CPPTRAJ and gistpp, for which I have been following the main tutorial at the AMBER website. However, I am still not clear on how to obtain the average number of water molecules per group as generated by “gistpp <…> -op group”.
>> One of the files generated by gist is called grcount.txt and at first sight it seems that it outputs the average number of molecules per group, but this is not the case, because the resulting values are too high (~90 for a single water molecule in the binding site). If I normalise against the length of the trajectory (1000 frames), then the values become too low (0.09 molecules, which is again not consistent with visually inspecting the trajectory).
>> Does anyone have experience obtaining the average number of water molecules per GIST group? Any help would be much appreciated.
>> Kind regards,Miro
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 04 2022 - 13:35:17 PDT
