Re: Help with "ERROR: Constraint failure in RATTLE algorithm for atom XXXX!"

From: Aron Broom (broomsday_at_gmail.com)
Date: Mon May 20 2013 - 14:07:56 CDT

tragic.

and you say that atom doesn't appear weird when visualizing? Are you
recording the DCD every timestep so you can catch the frame just before the
failure?

Maybe try being extreme in your energy minimization beforehand, as in
10,000 steps or something.

Also, are you gradually heating? I usually just initialize to 300K, but if
you are having problems, maybe do a slow heating of 1K per 1000 steps or
something. Are you using a Langevin thermostat?

On Mon, May 20, 2013 at 12:57 PM, Corey Fugate <coreyfugate_at_gmail.com>wrote:

> Good point. I commented out the lines in the configuration concerning
> periodic boundary conditions and PME. The run fails with the same error
> messages, but caused by different atoms.
>
>
> On Mon, May 20, 2013 at 10:16 AM, Aron Broom <broomsday_at_gmail.com> wrote:
>
>> you should not have a box if you are doing vacuum simulations.
>>
>> you need to not use periodic conditions and not use PME.
>>
>> ideally pairdist and cutoff would be essentially infinite.
>>
>> not sure if it is the problem, but when you use periodic conditions in a
>> vacuum, you have two issues:
>>
>> 1) PME doesn't work, because you don't have uniform density
>> 2) if your system has a net charge, you are simulating an infinite
>> lattice of similarly charged systems which is madness
>>
>> ~Aron
>>
>>
>>
>>
>> On Mon, May 20, 2013 at 12:08 PM, Corey Fugate <coreyfugate_at_gmail.com>wrote:
>>
>>> I am trying to minimize and equilibrate a protein monomer that consists
>>> of 6525 atoms (protein only). I am doing this is vacuum. I receive the
>>> following error in my log file -
>>>
>>> REINITIALIZING VELOCITIES AT STEP 100 TO 310 KELVIN
>>> TCL: Running for 2500 steps
>>> ERROR: Constraint failure in RATTLE algorithm for atom 1917!
>>> ERROR: Constraint failure; simulation has become unstable.
>>> ERROR: Exiting prematurely; see error messages above.
>>>
>>> This seems to be a common question on the mailing list, so I first read
>>> through some suggestions described previously. I've tried increasing the
>>> cutoff and the pairlistdist. This didn't seem to help. It did change the
>>> atom that seemed to cause the problem, it always seems to be to near the
>>> surface f the protein. I looked at the pdb in VMD and I don't see any
>>> unfavorable overlap between this atom and another. Jim Phillips mentioned
>>> that a good rule of thumb is 1 atom per 10 cubic angstroms. I am working
>>> with a box with periodic boundary conditions. If I have 6525 atoms, I
>>> would need a box that had a volume of 65,250 A^3. The original dimensions
>>> of my box set by the cellBasisVectors are 47,42,44. This would give me a
>>> box with an area of 80,856 A^3. I went ahead and make the box larger by
>>> making all the dimensions 50.
>>>
>>> Am I missing something obvious? Any suggestions would be great.
>>>
>>
>>
>>
>> --
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
>>
>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:14 CST