Re: Spherical restraints

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Tue Apr 06 2021 - 13:31:27 CDT

Hi Yasser, I can only speak for the logfile that you sent, but in that one
the MISC energy is quite high, about 14,000 kcal/mol. (Note that with
outputEnergy the energy of that restraint specifically would be written to
the colvars.traj file, which will make it easier to confirm).

Please do confirm that the initial configuration is really okay, and then
check again against a run with a CPU-only build on the cluster to make
sure. The restraint forces coming from Colvars will be identical: but how
NAMD will deal with them differs between the CPU and GPU builds.

ETITLE: TS BOND ANGLE DIHED IMPRP
              ELECT VDW BOUNDARY *MISC*
 KINETIC TOTAL TEMP POTENTIAL TOTAL3
     TEMPAVG PRESSURE GPRESSURE VOLUME
PRESSAVG GPRESSAVG

ENERGY: 0 235.1373 489.4649 479.7107 11.5964
        -55823.9645 5575.4001 0.0000 *14283.9594*
 9260.6370 -25488.0588 299.9977 -34748.6957 -25419.1273
      299.9977 49231.0264 49231.6119 166940.3348
49231.0264 49231.6119

If the initial configuration is fine and that "MISC" value is something
else and unrelated, then I would try to run a test with the colvar and its
restraint defined but with "forceConstant 0", or something very small.

Giacomo

On Tue, Apr 6, 2021 at 1:28 PM Almeida-Hernández, Yasser, Dr. <
yasser.almeida-hernandez_at_uni-due.de> wrote:

> Hi Giacomo,
>
>
> I was using NAMD_2.13_Linux-x86_64-multicore-CUDA for GPU,
> NAMD_2.14b2_Linux-x86_64-multicore for CPU, and a restraining force of
> 1000 kcal/mol.
>
>
> The initial configuration (the one I am running in CPU in my laptop) is
> fine a running. I tried now with NAMD_2.14b2_Linux-x86_64-multicore-CUDA
> in my cluster and the same happened with the -nan values (with
> outputEnergy = on). The MISC column in the output is 0.0000. So, no NAMD
> CUDA version for colvars.
>
>
> Then, I tweaked the configuration file a bit (after son errors), used
> NAMD_2.14b2_Linux-x86_64-multicore in my cluster (the same in my
> laptop), the run starts but stops after one step, with a segfault error.
> Attached are conf file, the log, and the slurm.out file.
>
>
> Yasser
> ------------------------------
> *From:* Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
> *Sent:* Tuesday, April 6, 2021 6:17:11 PM
> *To:* Almeida-Hernández, Yasser, Dr.
> *Cc:* NAMD list
> *Subject:* Re: namd-l: Spherical restraints
>
> Hi Yasser, which NAMD version and build are you running on the GPU?
> Colvars doesn't make use of the GPU, but only of CPU code. So the only way
> that the CPU and GPU versions may differ is if the forces applied by
> Colvars are exceedingly large, and the GPU NAMD build can't deal with them
> correctly.
>
> Is the initial configuration compatible with the restraint? You can add
> outputEnergy to the restraint, or (if that's the only custom force in your
> workflow) look at the MISC column of the NAMD output.
>
> Giacomo
>
> On Tue, Apr 6, 2021 at 11:50 AM Almeida-Hernández, Yasser, Dr. <
> yasser.almeida-hernandez_at_uni-due.de> wrote:
>
>> Hi Giacomo,
>>
>>
>> Thank you for your reply. Yes, that was the problem.
>>
>>
>> The setting looks now like:
>>
>>
>> set config "
>>
>> colvar {
>> name LIG_sphere
>> distance {
>> group1 {atomNumbers 704} # Cu2+ cation
>> group2 {atomNumbersRange 707-734} # Ligand
>> }
>> }
>>
>> harmonicWalls {
>> colvars LIG_sphere
>> outputEnergy on
>> lowerWalls 0
>> upperWalls 15
>> forceConstant 1000
>> }
>> "
>>
>> colvars on
>> cv config ${config}
>>
>>
>> Now, if I run it in my laptop it works well, but if I run it in a cluster
>> with GPU (namd2 +idlepoll +isomalloc_sync +p46 +devices 0,1 $INP.conf >
>> $INP.log), it writes -nan values:
>>
>>
>> TIMING: 89000 CPU: 686.545, 0.0172745/step Wall: 737.674,
>> 0.0173574/step, 4.39239 hours remaining, 672.355469 MB of memory in use.
>> ENERGY: 89000 nan 888064.2004 nan
>> nan nan 0.0000 0.0000
>> 0.0000 nan nan nan
>> nan -nan nan nan
>> nan nan nan nan
>>
>> Any thoughts?
>>
>> Best
>>
>> ------------------------------
>> *From:* Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
>> *Sent:* Tuesday, April 6, 2021 5:06:58 PM
>> *To:* NAMD list; Almeida-Hernández, Yasser, Dr.
>> *Subject:* Re: namd-l: Spherical restraints
>>
>> Hello Yasser, there are a few things to check, but the first one would be
>> to make sure that you actually loaded that configuration. There will be a
>> (fairly verbose) set of messages printed when you do.
>>
>> What you did there is setting a string variable "config" to contain the
>> configuration text, which can be used in a Tcl script but you actually need
>> to load it afterwards, e.g.:
>>
>> cv config ${config}
>>
>>
>>
>> On Tue, Apr 6, 2021 at 10:09 AM Almeida-Hernández, Yasser, Dr. <
>> yasser.almeida-hernandez_at_uni-due.de> wrote:
>>
>>> Hello,
>>>
>>>
>>> I want to spherically restrain a ligand around a Cu2+ cation with
>>> harmonic walls. For this, I am using the colvars module as:
>>>
>>>
>>> # Restraint the ligand in a sphere of r=15 AA around the Cu2+
>>> set config "
>>>
>>> colvars on
>>>
>>> colvar {
>>> name LIG_sphere
>>> distance {
>>> group1 {atomNumbers 704} # Cu2+ cation
>>> group2 {atomNumbersRange 707-734} # Ligand
>>> }
>>> }
>>>
>>> harmonicWalls {
>>> name LIG_sphere
>>> colvars HL
>>> outputEnergy on
>>> lowerWalls 0
>>> upperWalls 15
>>> forceConstant 1000
>>> }
>>> "
>>>
>>>
>>> ... but it doesn't work. The ligand tumbles all around.
>>>
>>>
>>> Am I missing something? Any help will be appreciated.
>>>
>>>
>>> Best,
>>>
>>> Yasser
>>>
>>>
>>> --
>>> Dr. Yasser Almeida-Hernandez
>>> Postdoctoral Researcher
>>> Computational Biochemistry group
>>> T01 R03 D45
>>> Faculty of Biology
>>> University of Duisburg-Essen
>>> Universitätsstr. 2, 45117 Essen
>>> Email: yasser.almeida-hernandez_at_uni-due.de
>>> <%20yasser.almeida-hernandez_at_uni-due.de>
>>> Phone: +49 201 183 2457 <+49%20201%20183%202457>
>>>
>>>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST