Re: [AMBER] huge reference.frc in MMPBSA

From: Ray Luo <rluo.uci.edu>
Date: Sun, 31 Mar 2019 16:51:03 -0400

Jubilant,

Could you email me, off the list, your input files? Please only keep
three snapshots in your trajectory file.

All the best,
Ray
--
Ray Luo, Ph.D.
Professor of Structural Biology/Biochemistry/Biophysics,
Chemical Physics, Biomedical Engineering, and Chemical Engineering
Department of Molecular Biology and Biochemistry
University of California, Irvine, CA 92697-3900
On Sun, Mar 31, 2019 at 1:46 PM Antoine MARION <amarion.metu.edu.tr> wrote:
>
> Dear Amber community,
>
> I came across an issue while using MMPBSA.py.
> This problem is related to a question raised last month by Jubilant so I attached the original email bellow.
>
> When using the default behavior of MMPBSA.py, the energy is calculated using “mmpbsa_py_energy”. An alternative is to set “use_sander=1” in the input to run the calculation by calling sander instead.
>
> Here is what happens for a trajectory of about 5000 frames (i.e., ~250MB without water):
> - with mmpbsa_py_energy (default):
> a huge file named “reference.frc” is created, in which forces on all atoms are printed for each processed frame.
> At the end of the calculation, the size of this file is 4.3 GB.
>
> - with sander (i.e., use_sander=1):
> the total size of the whole directory in which the calculation is performed is only about 500 MB big.
> This is more reasonable and expected as it represents about 250 MB for the trajectory of the complex, 250 MB for the receptor, and a few MB for that of the ligand.
>
>
> The problem becomes more serious when processing much larger trajectories. Namely, with our current setup (which might not be optimal yet I admit), the “reference.frc” file reaches up to 200 GB for an input trajectory of about 12 GB, which is an issue on some machines with limited user disk space.
>
> Running the calculation with “use_sander=1” is somehow a little bit slower for the test calculation briefly described above (about 1.5 times longer).
>
> I traced down the “reference.frc" file in the source and found it in “$AMBERHOME/AmberTools/src/sff/eff.c” where it is used to output forces for what seems to be a debugging purpose. Currently, the output of the forces is controlled in this routine by an option named “e_debug”, which also controls the output of the energy used by “MMPBSA.py”. I commented the corresponding lines and recompiled. The result is strictly the same and the total disk space usage is now back to about 500 MB. That tends to confort me in thinking that this file has no purpose for the MMPBSA calculation itself (at least when using gb).
>
> I would greatly appreciate if you could let me know if I overlooked something here and broke something else hidden by commenting those lines (the lines outputting the forces in “reference.frc”).
>
> If not and if this file is really not needed, would you consider changing the default behavior of the routine to avoid writing such a huge file?
>
> And finally, if my understanding of the problem is correct, I would say that the answer to Jubilant’s question should be that the file can be safely deleted.
>
> Thanks a lot in advance for your help.
>
> All the best,
> Antoine
>
>
> From: Sundar <jubilantsundar.gmail.com <mailto:jubilantsundar.gmail.com?Subject=Re%3A%20%5BAMBER%5D%20reference.frc%20file%20from%20MMPBSA%20calculations>>
> Date: Wed, 13 Feb 2019 12:02:55 -0600
>
> My mmpbsa calculations write a huge reference.frc file at the end of the
> calculation.
>
> Can I safely delete this file or it may be useful in any future
> calculations?
> Other files generated during the calculation were deleted at the end by
> default but not this.
> Any experienced input will be greatly appreciated.
>
> --
> Thanks,
> Jubilant
>
> ____________________________
> -- Dr. Antoine MARION
> Assistant Professor
> Doktor Öğretim Üyesi
>
> Department of Chemistry
> Middle East Technical University
> 06800, Ankara, Turkey
>
> Office: O-310
> Email: amarion.metu.edu.tr <mailto:amarion.metu.edu.tr>
> Phone: +90 312 210 51 44
>
> _______________________________________________
> 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 Sun Mar 31 2019 - 14:00:01 PDT
Custom Search