Re: [AMBER] MMPBSA results with Amber 12 versus Amber 16

From: Vlad Cojocaru <vlad.cojocaru.mpi-muenster.mpg.de>
Date: Fri, 14 Oct 2016 19:35:02 +0200

Hi Ray,

Now I don't have access to the computers but the differences I am seeing
in the deltaG are from -26 (amber12), to -40 (amber14), to -77 (amber16)
.... As the only difference in the outputs are in the EEL energy (all
other terms are identical), this is a change of 200% percent with each
Amber version ...

But I will take a look again at the test cases on Monday and try to
rationalize further ...

Its a little frustrating because I was quite happy with the values we
got using Amber 12 and now it turns out that simply changing the version
of the program we use, we get huge differences to our original results
... Maybe the ddGs or dddGs we reported in the Structure paper 2014 are
unaffected (did not do a full test for that yet) but still it is
frustrating to see such differences by just changing the version of the
program .....

Best
Vlad

On 10/14/2016 06:34 PM, Ray Luo wrote:
> Hi Vlad,
>
> I found the reason of the discrepancy. If you look at the standard
> nonlinear PB test cases (under AmberTools/test/pbsa_npb), they are
> consistent between amber12 and amber14. While working on his NPB/P3M
> algorithm for the amber16 release, my student, Li Xiao, found a flip
> of sign in the bulk ion energy calculation. You can find his comment
> in routine pb_ionene() within "pb_nlsolver.F90". After his fix, we see
> a change about 0.7% in the total EELEC energies in our test cases.
> I'll confirm with him again regarding this.
>
> All the best,
> Ray
> --
> Ray Luo, Ph.D.
> Professor
> Biochemistry, Molecular Biophysics, Chemical Physics,
> Chemical and Biomedical Engineering
> University of California, Irvine, CA 92697-3900
>
>
> On Fri, Oct 14, 2016 at 8:44 AM, Ray Luo <rluo.uci.edu> wrote:
>> Vlad,
>>
>> As for the P3M approach, you probably don't want to freely adjust the
>> cutoff distance because the nonbonded list is shared between direct
>> pairwise sum and surface generation routines.
>>
>> I'm looking at your input files right now. In general, we have mostly
>> focussed on the default setup that is used by most users to ensure
>> consistency between releases. Other rarely used options are not tested
>> often. Maybe we should put your test into the release test cases so we
>> can check it every time the code is changed. What do you think?
>>
>> All the best,
>> Ray
>> --
>> Ray Luo, Ph.D.
>> Professor
>> Biochemistry, Molecular Biophysics, Chemical Physics,
>> Chemical and Biomedical Engineering
>> University of California, Irvine, CA 92697-3900
>>
>>
>> On Fri, Oct 14, 2016 at 1:42 AM, Vlad Cojocaru
>> <vlad.cojocaru.mpi-muenster.mpg.de> wrote:
>>> Thanks ...
>>>
>>> Sorry, I need to ask another question ... I am now playing with lots of
>>> parameters and in one run I set cutnb to a higher value (with eneopt =
>>> 4) but I get an error that points me to a "cutres" option which I could
>>> not find in the manual.
>>>
>>> Could you please let me know me what the "cutres" option is ?
>>>
>>> Thanks
>>> Vlad
>>>
>>> On 10/14/2016 12:23 AM, Ray Luo wrote:
>>>> I'm teaching this quarter and haven't got a time to look at your
>>>> files, but I'll look at your example next ...
>>>>
>>>> It's most likely due to the different default optimal values and/or
>>>> bug fixes ... I will run your jobs without specifying any parameters
>>>> first to see whether the default behaviors are similar. This is how
>>>> new releases were first tested against previous versions. When you
>>>> specify most not not all parameters explicitly, the default behaviors
>>>> would get changed somewhat.
>>>>
>>>> As for speedup in Amber16, yes, this is due to a major code cleanup to
>>>> organize the modules better.
>>>>
>>>> All the best,
>>>> Ray
>>>> --
>>>> Ray Luo, Ph.D.
>>>> Professor
>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>> Chemical and Biomedical Engineering
>>>> University of California, Irvine, CA 92697-3900
>>>>
>>>>
>>>> On Thu, Oct 13, 2016 at 8:43 AM, Vlad Cojocaru
>>>> <vlad.cojocaru.mpi-muenster.mpg.de> wrote:
>>>>> Dear Ray, Dear all,
>>>>>
>>>>> I also performed exactly the same calculation I described in my original
>>>>> mail with Amber 14 .. Every version of Amber provides different values for
>>>>> the electrostatic energies although the run was done with exactly the same
>>>>> scripts, on exactly same topologies and trajectory .. Moreover, almost all
>>>>> pbsa parameters are specified explicitly in the input MDIN file (see again
>>>>> original mail below) ...
>>>>>
>>>>> Best wishes
>>>>> Vlad
>>>>>
>>>>>
>>>>>
>>>>> On 10/13/2016 10:47 AM, Vlad Cojocaru wrote:
>>>>>> Hi Ray,
>>>>>>
>>>>>> It would be great if we could understand where these differences come
>>>>>> from ...I looked at all parameters from my customized MDIN file (which I
>>>>>> attached to my original mail below) and I did not notice any change
>>>>>> between Amber 12 and Amber 16 in terms of the meaning of the values
>>>>>> (hope I did not miss anything). I am now doing the same calculation
>>>>>> with Amber 14 as well to see if the difference came between Amber 12 and
>>>>>> Amber 14 or between Amber 14 to Amber 16 ...
>>>>>>
>>>>>> On a side note, the calculation is much faster with Amber 16 comparing
>>>>>> to Amber 14 ... Is that to be expected ?
>>>>>>
>>>>>> Thanks for looking into this
>>>>>> Vlad
>>>>>>
>>>>>> On 10/12/2016 10:44 AM, Ray Luo wrote:
>>>>>>> Hi Vlad,
>>>>>>>
>>>>>>> Thanks a lot for letting us know! I suppose the default was changed in
>>>>>>> the script or the code. Will let you know the cause.
>>>>>>>
>>>>>>> All the best,
>>>>>>> Ray
>>>>>>> --
>>>>>>> Ray Luo, Ph.D.
>>>>>>> Professor
>>>>>>> Biochemistry, Molecular Biophysics, Chemical Physics,
>>>>>>> Chemical and Biomedical Engineering
>>>>>>> University of California, Irvine, CA 92697-3900
>>>>>>>
>>>>>>>
>>>>>>> On Wed, Oct 12, 2016 at 1:31 AM, Vlad Cojocaru
>>>>>>> <vlad.cojocaru.mpi-muenster.mpg.de> wrote:
>>>>>>>> Dear all,
>>>>>>>>
>>>>>>>> I am trying to reproduce some of our previous MMPBSA calculations with
>>>>>>>> Amber
>>>>>>>> 16. Original calculations were done in Amber 12. Now, using exactly the
>>>>>>>> same
>>>>>>>> trajectory, exactly same topology files, exactly the same input files
>>>>>>>> (see
>>>>>>>> below the MMPBSA input as well as the customized MDIN), I get an
>>>>>>>> absolute
>>>>>>>> affinity (without entropy) of -77 kcal/mol versus previously calculated
>>>>>>>> -26
>>>>>>>> kcal/mol ... The only difference between the runs in actually in the
>>>>>>>> electrostatic energy (-43.5 in the new calculation versus +8 in the old
>>>>>>>> calculation) . See attached output files.
>>>>>>>>
>>>>>>>> Therefore, the problem (or difference) is in the PB solver in Amber 16
>>>>>>>> versus Amber 12 ... Does anybody have any idea where the difference
>>>>>>>> could
>>>>>>>> come from ??
>>>>>>>>
>>>>>>>> I know that one should not put too much weight on the absolute values,
>>>>>>>> but
>>>>>>>> still running with exactly the same scripts, exactly same topology
>>>>>>>> files,
>>>>>>>> exactly the same old trajectory in 2 different versions of the same
>>>>>>>> program
>>>>>>>> should give the same results ....
>>>>>>>>
>>>>>>>> Thanks for any insights in this
>>>>>>>>
>>>>>>>> Best wishes
>>>>>>>> Vlad
>>>>>>>>
>>>>>>>>
>>>>>>>> ---- MMPBSA INPUT -----
>>>>>>>> MMPBSA
>>>>>>>> &general
>>>>>>>> debug_printlevel=1,
>>>>>>>> startframe=${startframe},
>>>>>>>> endframe=${endframe},
>>>>>>>> interval=${interval},
>>>>>>>> keep_files=1,
>>>>>>>> netcdf=1,
>>>>>>>> ligand_mask=":${r1_ligand}-${r2_ligand}",
>>>>>>>> receptor_mask=":${r1_receptor}-${r2_receptor}",
>>>>>>>> use_sander=1,
>>>>>>>> entropy=0,
>>>>>>>> full_traj=1,
>>>>>>>> verbose=2,
>>>>>>>> /
>>>>>>>> &pb
>>>>>>>> inp=2,
>>>>>>>> cavity_offset=-0.5692,
>>>>>>>> cavity_surften=0.0378,
>>>>>>>> indi=4.0,
>>>>>>>> exdi=80.0,
>>>>>>>> fillratio=4.0,
>>>>>>>> istrng=0.100,
>>>>>>>> linit=1000,
>>>>>>>> prbrad=1.4,
>>>>>>>> radiopt=1,
>>>>>>>> scale=2.0,
>>>>>>>> /
>>>>>>>>
>>>>>>>> $MPI_HOME/bin/mpirun -n $NSLOTS $AMBERHOME/bin/MMPBSA.py.MPI -O -i
>>>>>>>> mmpbsa_${run}.in \
>>>>>>>> -o
>>>>>>>> mmpbsa_${run}.out \
>>>>>>>> -cp
>>>>>>>> ${top_complex} \
>>>>>>>> -rp
>>>>>>>> ${top_receptor} \
>>>>>>>> -lp
>>>>>>>> ${top_ligand} \
>>>>>>>> -y
>>>>>>>> ${traj_complex} \
>>>>>>>> -eo
>>>>>>>> energy_${run}.out \
>>>>>>>> -use-mdins
>>>>>>>>
>>>>>>>>
>>>>>>>> ---- MDIN ------------------------
>>>>>>>> MMPBSA, Nonlinear PB, inp=2, sasopt=2
>>>>>>>> &cntrl
>>>>>>>> nsnb=99999, dec_verbose=0, ioutfm=1,
>>>>>>>> ipb=2, ntb=0, cut=999.0, imin=5,
>>>>>>>> igb=10, inp=2,
>>>>>>>> /
>>>>>>>> &pb
>>>>>>>> epsin=4, epsout=80, smoothopt=1,
>>>>>>>> istrng=100.0, pbtemp=300, radiopt=1,
>>>>>>>> dprob=1.4, iprob=2.0, sasopt=2, saopt=0,
>>>>>>>> triopt=1, arcres=0.25,
>>>>>>>> npbopt=1, solvopt=1, accept=0.001,
>>>>>>>> maxitn=100, fillratio=4.0, space=0.5,
>>>>>>>> nbuffer=0, nfocus=2, fscale=8, npbgrid=1,
>>>>>>>> bcopt=5, eneopt=1, frcopt=0, scalec=0,
>>>>>>>> cutfd=5.0, cutnb=12, nsnba=1,
>>>>>>>> phiout=0,
>>>>>>>> decompopt=2, use_rmin=1, sprob=0.557, vprob=1.3,
>>>>>>>> rhow_effect=1.129, use_sav=1,
>>>>>>>> cavity_surften=0.0378, cavity_offset=-0.5692,
>>>>>>>> maxsph=400,
>>>>>>>> /
>>>>>>>>
>>>>>>>>
>>>>>>>> --
>>>>>>>> Dr. Vlad Cojocaru
>>>>>>>> Computational Structural Biology Laboratory
>>>>>>>> Department of Cell and Developmental Biology
>>>>>>>> Max Planck Institute for Molecular Biomedicine
>>>>>>>> Röntgenstrasse 20, 48149 Münster, Germany
>>>>>>>> Tel: +49-251-70365-324; Fax: +49-251-70365-399
>>>>>>>> Email: vlad.cojocaru[at]mpi-muenster.mpg.de
>>>>>>>> http://www.mpi-muenster.mpg.de/43241/cojocaru
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> 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
>>>>> --
>>>>> Dr. Vlad Cojocaru
>>>>> Computational Structural Biology Laboratory
>>>>> Department of Cell and Developmental Biology
>>>>> Max Planck Institute for Molecular Biomedicine
>>>>> Röntgenstrasse 20, 48149 Münster, Germany
>>>>> Tel: +49-251-70365-324; Fax: +49-251-70365-399
>>>>> Email: vlad.cojocaru[at]mpi-muenster.mpg.de
>>>>> http://www.mpi-muenster.mpg.de/43241/cojocaru
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> 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
>>> --
>>> Dr. Vlad Cojocaru
>>> Computational Structural Biology Laboratory
>>> Department of Cell and Developmental Biology
>>> Max Planck Institute for Molecular Biomedicine
>>> Röntgenstrasse 20, 48149 Münster, Germany
>>> Tel: +49-251-70365-324; Fax: +49-251-70365-399
>>> Email: vlad.cojocaru[at]mpi-muenster.mpg.de
>>> http://www.mpi-muenster.mpg.de/43241/cojocaru
>>>
>>>
>>> _______________________________________________
>>> 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

-- 
Dr. Vlad Cojocaru
Computational Structural Biology Laboratory
Department of Cell and Developmental Biology
Max Planck Institute for Molecular Biomedicine
Röntgenstrasse 20, 48149 Münster, Germany
Tel: +49-251-70365-324; Fax: +49-251-70365-399
Email: vlad.cojocaru[at]mpi-muenster.mpg.de
http://www.mpi-muenster.mpg.de/43241/cojocaru
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Oct 14 2016 - 11:00:03 PDT
Custom Search