[gmx-users] 1-4 interaction between X and Y at distance 1.804 which is larger than 1-4 table size 1.800 nm.

Mark Abraham Mark.Abraham at anu.edu.au
Thu Jan 20 00:07:57 CET 2011


On 20/01/2011 9:53 AM, TJ Mustard wrote:
>
>
> On January 18, 2011 at 12:50 PM Mark Abraham <Mark.Abraham at anu.edu.au> 
> wrote:
>
>> On 19/01/2011 7:33 AM, TJ Mustard wrote:
>>>
>>>
>>> On January 17, 2011 at 3:48 PM Mark Abraham 
>>> <Mark.Abraham at anu.edu.au> <mailto:Mark.Abraham at anu.edu.au> wrote:
>>>
>>>> On 18/01/2011 10:27 AM, TJ Mustard wrote:
>>>>>
>>>>>
>>>>> On January 17, 2011 at 2:47 PM TJ Mustard <mustardt at onid.orst.edu> 
>>>>> <mailto:mustardt at onid.orst.edu> wrote:
>>>>>
>>>>>> On January 17, 2011 at 2:21 PM "Justin A. Lemkul" 
>>>>>> <jalemkul at vt.edu> <mailto:jalemkul at vt.edu> wrote:
>>>>>>
>>>>>> >
>>>>>> >
>>>>>> > TJ Mustard wrote:
>>>>>> > >
>>>>>> > >
>>>>>> > > No this is a completely different error that occurs when the 
>>>>>> other (EM)
>>>>>> > > error does not.
>>>>>> > >
>>>>>> >
>>>>>> > So, EM completes, but you then get this unstable system?  How 
>>>>>> well did EM
>>>>>> > converge?  Were the potential and maximum force acceptable?
>>>>>> >
>>>>>> > Instabilities are generally caused by incorrect topologies or 
>>>>>> .mdp settings.
>>>>>> > Those are the first places I would look for sources of error.
>>>>>> >
>>>>>> > -Justin
>>>>>> >
>>>>>> > >
>>>>>> > >
>>>>>> > > Thank you,
>>>>>> > >
>>>>>> > > TJ Mustard
>>>>>> > >
>>>>>> > > On January 17, 2011 at 1:51 PM "Justin A. Lemkul" 
>>>>>> <jalemkul at vt.edu> <mailto:jalemkul at vt.edu> wrote:
>>>>>> > >
>>>>>> > > >
>>>>>> > > >
>>>>>> > > > TJ Mustard wrote:
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > > Hi all,
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > > We have been running a lot of simulations on protein ligand
>>>>>> > > > > interactions, and my settings/setup/mdp files worked 
>>>>>> great for one
>>>>>> > > > > system. Then when we moved to a larger and more 
>>>>>> complicated system we
>>>>>> > > > > started getting mdrun 1-4 interaction errors in our MD.
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > > We are running with heavyh and have correctly made our 
>>>>>> ligands with
>>>>>> > > > > acpype and edited to have heavyh.
>>>>>> > > > >
>>>>>> > > > > Same setup as our smaller simpler model, but keep getting 
>>>>>> these:
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > > Warning: 1-4 interaction between X and Y at distance 
>>>>>> 1.804 which is
>>>>>> > > > > larger than 1-4 table size 1.800 nm.
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > > > The jobs will continue on and finish but we are worried 
>>>>>> about the error
>>>>>> > > > > that could be included with these errors.
>>>>>> > > > >
>>>>>> > > > >
>>>>>> > > >
>>>>>> > > > Is this related to the other issue we're discussing?  If 
>>>>>> so, please
>>>>>> > > do not
>>>>>> > > > double-post, especially if you're including different 
>>>>>> information in
>>>>>> > > different
>>>>>> > > > threads.  It becomes a nightmare to try to keep track of.
>>>>>> > > >
>>>>>> > > > That particular error is usually not a good sign.  Normal, 
>>>>>> stable
>>>>>> > > simulations
>>>>>> > > > should not give such an error.
>>>>>> > > >
>>>>>> > > > -Justin
>>>>>> > > >
>>>>>>
>>>>>> Here is the tail of my em:
>>>>>>
>>>>>> Step= 3438, Dmax= 2.5e-04 nm, Epot= -1.05878e+06 Fmax= 
>>>>>> 8.91784e+01, atom= 43563
>>>>>> Step= 3439, Dmax= 3.0e-04 nm, Epot= -1.05878e+06 Fmax= 
>>>>>> 8.89898e+01, atom= 43563
>>>>>> Step= 3440, Dmax= 3.6e-04 nm, Epot= -1.05878e+06 Fmax= 
>>>>>> 3.27390e+02, atom= 24
>>>>>> Step= 3441, Dmax= 4.3e-04 nm, Epot= -1.05878e+06 Fmax= 
>>>>>> 3.34235e+02, atom= 24
>>>>>> Step= 3451, Dmax= 1.0e-06 nm, Epot= -1.05878e+06 Fmax= 
>>>>>> 3.32604e+02, atom= 24563
>>>>>> Stepsize too small, or no change in energy.
>>>>>> Converged to machine precision,
>>>>>> but not to the requested precision Fmax < 20
>>>>>>
>>>>>> Double precision normally gives you higher accuracy.
>>>>>> You might need to increase your constraint accuracy, or turn
>>>>>> off constraints alltogether (set constraints = none in mdp file)
>>>>>>
>>>>>> writing lowest energy coordinates.
>>>>>>
>>>>>> Steepest Descents converged to machine precision in 3452 steps,
>>>>>> but did not reach the requested Fmax < 20.
>>>>>> Potential Energy  = -1.0587850e+06
>>>>>> Maximum force     =  3.3423511e+02 on atom 24
>>>>>> Norm of force     =  4.9464898e+00
>>>>>>
>>>>>> My major settings are:
>>>>>>
>>>>>> define                   = -DPOSRES_WATER
>>>>>>
>>>>>> integrator               = steep
>>>>>>
>>>>>> nsteps                   = 10000
>>>>>>
>>>>>> emtol                    = 20
>>>>>> emstep                   = 0.01
>>>>>>
>>>>>> Am I setting my emtol to low? I occasionally see a Fmax of 
>>>>>> ~250-900 but mostly above that.
>>>>>>
>>>>>> I am currently having the epiphany that -DPOSRES_WATER could be 
>>>>>> my problem.
>>>>>>
>>>>>> Thank you,
>>>>>>
>>>>>> TJ Mustard
>>>>>>
>>>>> OK I ran the EM without the -DPOSRES_WATER defined and continued 
>>>>> through my script to pr and md, and I got this error on my first 
>>>>> MD after a positional restraint md (we call it a PR).
>>>>>
>>>>> First the EM tail:
>>>>>
>>>>> Step= 5519, Dmax= 8.3e-04 nm, Epot= -1.09261e+06 Fmax= 
>>>>> 8.13679e+02, atom= 24
>>>>> Step= 5520, Dmax= 1.0e-03 nm, Epot= -1.09261e+06 Fmax= 
>>>>> 7.21335e+02, atom= 24
>>>>> Step= 5523, Dmax= 3.0e-04 nm, Epot= -1.09261e+06 Fmax= 
>>>>> 2.58642e+02, atom= 24
>>>>> Step= 5527, Dmax= 4.5e-05 nm, Epot= -1.09261e+06 Fmax= 
>>>>> 1.88718e+02, atom= 24936
>>>>> Step= 5533, Dmax= 1.7e-06 nm, Epot= -1.09261e+06 Fmax= 
>>>>> 1.86052e+02, atom= 24
>>>>> Stepsize too small, or no change in energy.
>>>>> Converged to machine precision,
>>>>> but not to the requested precision Fmax < 20
>>>>>
>>>>> Double precision normally gives you higher accuracy.
>>>>> You might need to increase your constraint accuracy, or turn
>>>>> off constraints alltogether (set constraints = none in mdp file)
>>>>>
>>>>> writing lowest energy coordinates.
>>>>>
>>>>> Steepest Descents converged to machine precision in 5534 steps,
>>>>> but did not reach the requested Fmax < 20.
>>>>> Potential Energy  = -1.0926141e+06
>>>>> Maximum force     =  1.8871829e+02 on atom 24
>>>>> Norm of force     =  4.8353853e+00
>>>>>
>>>>
>>>> That looks fine.
>>>>
>>>>> Now the MD tail:
>>>>>
>>>>> No previous checkpoint file present, assuming this is a new run.
>>>>>
>>>>> Back Off! I just backed up md.log to ./#md.log.1#
>>>>> Getting Loaded...
>>>>> Reading file RFP-t_md.tpr, VERSION 4.5.3 (single precision)
>>>>> Starting 2 threads
>>>>> Loaded with Money
>>>>>
>>>>> Making 1D domain decomposition 2 x 1 x 1
>>>>> starting mdrun 'ligand in water'
>>>>> 25000 steps,    100.0 ps.
>>>>> step 10400, will finish Mon Jan 17 15:51:43 2011Warning: 1-4 
>>>>> interaction between 108 and 116 at distance 1.800 which is larger 
>>>>> than the 1-4 table size 1.800 nm
>>>>> These are ignored for the rest of the simulation
>>>>> This usually means your system is exploding,
>>>>> if not, you should increase table-extension in your mdp file
>>>>> or with user tables increase the table size
>>>>> step 11700, will finish Mon Jan 17 15:52:10 2011imb F  0%
>>>>>
>>>>> Any thoughts? Remember this system is doing FEP and heavy hydrogens.
>>>>>
>>>>
>>>> Simplify things while troubleshooting. Does normal MD work instead 
>>>> of FEP? Do normal hydrogens work?
>>>>
>>>> Use nstxout = 1 and look and the region around atoms 108 116 to see 
>>>> what is breaking. That may tell you why it is breaking.
>>>>
>>>>> But so was the positional restraint md, and it did not error.
>>>>>
>>>>
>>>> Shrug, numerical instabilities can take time to show up.
>>>>
>>>> Mark
>>>
>>> I have ran both light hydrogen with and without FEP on and yes the 
>>> runs with the FEP give the 1-4 interaction errors. I will soon be 
>>> running Heavy w/o FEP. The other very interesting part is that 108 
>>> and 116 are 17.2 Angstroms apart. They are on the same 
>>> molecule(drug) but very far apart and definitely not bonded.
>>>
>>
>> Something's impossible there. 1-4 interactions can arise only when 
>> three bonds connect the atoms. What's your drug topology?
>>
>>> Also my rvdw, rlist, and rcoulomb are all 0.8, which is to my 
>>> knowledge 8 angstroms. How are two particles, though on the same 
>>> molecule, interacting outside the PME/Cut-off distance?
>>>
>>
>> 1-4 interactions have nothing to do with PME or cut-offs. See manual 
>> 4.6.1.
>>
>> Also, such short cut-offs need compensating changes in ewald_rtol and 
>> fourier_spacing. 1.0nm with 1e-5 and 0.12nm are normal, well-tested 
>> values.
>>>
>>> And if this is a problem do I need to increase my PME/Cut-off 
>>> distance such that the entire molecule is within that distance. I 
>>> would rather not, due to speed issues.
>>>
>>
>> Certainly not, even if it was relevant :-)
>>
>> Mark
>
> Mark and all,
>
> This
>

To what does "this" refer?

> has fixed the 1-4 interaction errors but has cause dynamic load 
> balancing to be turned on. What can be done to, other than smaller 
> pme/cut-offs to limit dynamic load balancing? While still keeping the 
> precision of course.
>

See mdrun -h for the dynamic load-balancing control. But why do you want 
to turn it off? Performance is better with it on.

Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20110120/3e4f291e/attachment.html>


More information about the gromacs.org_gmx-users mailing list