[gmx-users] genconf and bonded interactions

Wed Jul 27 15:49:04 CEST 2011

On 07/27/2011 03:38 PM, Justin A. Lemkul wrote:
>
> Please make sure to keep the discussion on the list.

Sorry, I just was continuing the discussion after sending you my 
coordinates.

>
> Fabio Affinito wrote:
>> Justin,
>> first of all, I thank you very much for your help: it's very precious.
>> I regret I didn't notice before the molecule was broken.
>
> I suggested to you yesterday that your initial configuration was broken.
> You told me it wasn't.

I checked with VMD but I was in error. I acknowledged my error, I think. 
Didn't I?

>
>> Unfortunately the problem is not solved with your recipe. I used
>> trjconv on the starting configuration and then I used the genconf.
>> The mdrun fails with the same error. In this case, the atoms involved
>> in the error are not the same of the previous one. But once again the
>> bond length is completely odd.
>> During the grompp I had some warning, but I can't understand where
>> they are coming from:
>>
>
> Certain bonded interactions can't take place between massless particles.
> As the warning states, if pdb2gmx produced the topology, don't worry
> about it.
>
>>> Cleaning up constraints and constant bonded interactions with virtual
>>> sites
>>> Removed 1305 Angles with virtual sites, 6251 left
>>> Removed 480 Proper Dih.s with virtual sites, 363 left
>>> Converted 2230 Constraints with virtual sites to connections, 2596 left
>>> Warning: removed 582 Constraints with vsite with Virtual site 3out
>>> construction
>>> This vsite construction does not guarantee constant bond-length
>>> If the constructions were generated by pdb2gmx ignore this warning
>>> Cleaning up constraints and constant bonded interactions with virtual
>>> sites
>>> Removed 1305 Angles with virtual sites, 6251 left
>>> Removed 480 Proper Dih.s with virtual sites, 363 left
>>> Converted 2230 Constraints with virtual sites to connections, 2596 left
>>> Warning: removed 582 Constraints with vsite with Virtual site 3out
>>> construction
>>> This vsite construction does not guarantee constant bond-length
>>> If the constructions were generated by pdb2gmx ignore this warning
>>> Cleaning up constraints and constant bonded interactions with virtual
>>> sites
>>> Removed 1305 Angles with virtual sites, 6251 left
>>> Removed 480 Proper Dih.s with virtual sites, 363 left
>>> Converted 2230 Constraints with virtual sites to connections, 2596 left
>>> Warning: removed 582 Constraints with vsite with Virtual site 3out
>>> construction
>>> This vsite construction does not guarantee constant bond-length
>>> If the constructions were generated by pdb2gmx ignore this warning
>>> Cleaning up constraints and constant bonded interactions with virtual
>>> sites
>>> Removed 1305 Angles with virtual sites, 6251 left
>>> Removed 480 Proper Dih.s with virtual sites, 363 left
>>> Converted 2230 Constraints with virtual sites to connections, 2596 left
>>> Warning: removed 582 Constraints with vsite with Virtual site 3out
>>> construction
>>> This vsite construction does not guarantee constant bond-length
>>> If the constructions were generated by pdb2gmx ignore this warning
>>
>> According to your experience, this can be pertinent with my problems?
>>
>
> Not likely. The virtual sites are within the protein, not the lipids,
> right?
>
> Without the mdrun output, there's little else I can tell you.

This is the mdrun output:

Initializing Domain Decomposition on 4096 nodes
Dynamic load balancing: auto
Will sort the charge groups at every domain (re)decomposition
Initial maximum inter charge-group distances:
     two-body bonded interactions: 30.031 nm, LJ-14, atoms 40702 40705
   multi-body bonded interactions: 30.031 nm, Angle, atoms 40701 40704
Minimum cell size due to bonded interactions: 33.035 nm
Maximum distance for 7 constraints, at 120 deg. angles, all-trans: 1.139 nm
Estimated maximum distance required for P-LINCS: 1.139 nm
Guess for relative PME load: 0.43
Will use 2400 particle-particle and 1696 PME only nodes
This is a guess, check the performance at the end of the log file
Using 1696 separate PME nodes
Scaling the initial minimum size with 1/0.8 (option -dds) = 1.25
Optimizing the DD grid for 2400 cells with a minimum initial size of 
41.293 nm
The maximum allowed number of cells is: X 0 Y 0 Z 0

-------------------------------------------------------
Program mdrun_mpi_bg, VERSION 4.5.4
Source code file: domdec.c, line: 6438

Fatal error:
There is no domain decomposition for 2400 nodes that is compatible with 
the given box and a minimum cell size of 41.2932 nm
Change the number of nodes or mdrun option -rdd or -dds
Look in the log file for details on the domain decomposition
For more information and tips for troubleshooting, please check the GROMACS
website at http://www.gromacs.org/Documentation/Errors
-------------------------------------------------------

Fabio

>
> -Justin
>
>> Thank you once again,
>>
>> Fabio
>>
>> On 07/27/2011 02:45 PM, Justin A. Lemkul wrote:
>>>
>>> The problem is indeed as I suspected. Your molecules are broken across
>>> periodic boundaries in the initial configuration. When you replicate
>>> with genbox, you've now replicated broken molecules such that those at
>>> the new interfaces between the boxes appear to be intact, but in fact
>>> they are not. This is just an artifact of visualization. If you use
>>> "trjconv -pbc whole" on the original starting configuration to make the
>>> molecules whole before replicating with genconf -nbox, you will avoid
>>> this problem.
>>>
>>> -Justin
>>>
>>> Fabio Affinito wrote:
>>>> Yes. It's in attachment.
>>>>
>>>> Fabio
>>>>
>>>> On 07/27/2011 02:38 PM, Justin A. Lemkul wrote:
>>>>>
>>>>> Can you please also send me the coordinate file from before you ran
>>>>> genconf?
>>>>>
>>>>> -Justin
>>>>>
>>>>> Fabio Affinito wrote:
>>>>>> Finally, I just got the right compression to fit my file in an
>>>>>> attachment.
>>>>>> The interesting lines are:
>>>>>> 159988 and 159992
>>>>>>
>>>>>> Kindly regards,
>>>>>>
>>>>>> Fabio
>>>>>>
>>>>>> On 07/27/2011 01:42 PM, Justin A. Lemkul wrote:
>>>>>>>
>>>>>>>
>>>>>>> Fabio Affinito wrote:
>>>>>>>> On 07/26/2011 07:02 PM, Justin A. Lemkul wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Fabio Affinito wrote:
>>>>>>>>>> On 07/26/2011 05:06 PM, Justin A. Lemkul wrote:
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Fabio Affinito wrote:
>>>>>>>>>>>> Maybe this is a different issue... but it's ok that after the
>>>>>>>>>>>> 99,999th
>>>>>>>>>>>> atom the counter restarts from zero?
>>>>>>>>>>>>
>>>>>>>>>>>>> 21374SOL OW99999 12.986 9.021 7.036 -0.0037 -0.4345 0.3977
>>>>>>>>>>>>> 21374SOL HW1 0 13.069 8.987 7.081 0.5916 0.5409 0.0638
>>>>>>>>>>>>
>>>>>>>>>>>> Could this be the origin of my problem?
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Atom numbering is not the problem. This happens all the time for
>>>>>>>>>>> systems
>>>>>>>>>>> of hundreds of thousands of atoms, which Gromacs handles just
>>>>>>>>>>> fine.
>>>>>>>>>>> Please investigate the points I suggested before.
>>>>>>>>>>
>>>>>>>>>> Yes, but this doesn't make things easier! :-)
>>>>>>>>>> According to the log the atoms to consider are 159986 and 159990
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> That's not what you posted before. The .log output indicated atoms
>>>>>>>>> 193657 and 193660 were problematic.
>>>>>>>>
>>>>>>>> Sorry. It's because I've tried with many systems (with different
>>>>>>>> -nbox
>>>>>>>> values) and the error was always the same.
>>>>>>>>
>>>>>>>>>
>>>>>>>>>> Browsing the conf.gro, if I didn't make mistakes this atoms are:
>>>>>>>>>>
>>>>>>>>>>> 30040POPG OE59986 0.080 13.158 2.964 0.0885 0.4154 -0.0859
>>>>>>>>>>
>>>>>>>>>>> 30040POPG C1C59990 0.219 26.034 3.221 0.6449 0.0750 0.1313
>>>>>>>>>>
>>>>>>>>>> But their distance is 12.8nm, while md.log reports 38.911 nm...
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>> In any case, why are atoms four bonds (based on the original .log
>>>>>>>>> output
>>>>>>>>> of 1-4 interactions being a problem) away separated by 12.8 nm?
>>>>>>>>> Seems
>>>>>>>>> very odd to me. I ask yet again - what are your box vectors,
>>>>>>>>> before
>>>>>>>>> and
>>>>>>>>> after manipulation with genconf?
>>>>>>>>
>>>>>>>>
>>>>>>>> Seems odd to me, too. For the box vectors (sorry if I didnt answer
>>>>>>>> before), in this case:
>>>>>>>>
>>>>>>>> after: 26.04658 26.04658 8.75317
>>>>>>>> before: 13.02329 13.02329 8.75317
>>>>>>>>
>>>>>>>> Hope it helps.
>>>>>>>>
>>>>>>>
>>>>>>> Without seeing the coordinate files, I can't tell anything
>>>>>>> further. If
>>>>>>> you want, you can send me (off-list) the coordinate files before and
>>>>>>> after genconf and please tell me the exact atoms that are
>>>>>>> reported as
>>>>>>> problematic.
>>>>>>>
>>>>>>> -Justin
>>>>>>>
>>>>>>>> Best regards,
>>>>>>>>
>>>>>>>> Fabio
>>>>>>>>
>>>>>>>>>
>>>>>>>>> -Justin
>>>>>>>>>
>>>>>>>>>> So what?
>>>>>>>>>>
>>>>>>>>>> F.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> -Justin
>>>>>>>>>>>
>>>>>>>>>>>> Thanks again,
>>>>>>>>>>>>
>>>>>>>>>>>> Fabio
>>>>>>>>>>>>
>>>>>>>>>>>> On 07/26/2011 04:38 PM, Justin A. Lemkul wrote:
>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> Fabio Affinito wrote:
>>>>>>>>>>>>>> On 07/26/2011 04:30 PM, Justin A. Lemkul wrote:
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Fabio Affinito wrote:
>>>>>>>>>>>>>>>> On 07/26/2011 04:19 PM, Justin A. Lemkul wrote:
>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> Were the molecules whole in the coordinate file you
>>>>>>>>>>>>>>>>> replicated? If
>>>>>>>>>>>>>>>>> not,
>>>>>>>>>>>>>>>>> the bonds will now be assigned across the entire box.
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>>> -Justin
>>>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>>> Yes and not, depending on what you mean by "whole".
>>>>>>>>>>>>>>>> It is an ion channel, so it's made of four chains.
>>>>>>>>>>>>>>>> This clarifies? (i guess not..)
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> By whole, I mean that the molecules are not split across
>>>>>>>>>>>>>>> periodic
>>>>>>>>>>>>>>> boundaries in the initial configuration that you
>>>>>>>>>>>>>>> replicated. If
>>>>>>>>>>>>>>> you
>>>>>>>>>>>>>>> replicate a periodic break, then you split the molecules
>>>>>>>>>>>>>>> by a
>>>>>>>>>>>>>>> distance
>>>>>>>>>>>>>>> equal to the new periodic distance.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> -Justin
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>> Ok, so: no, it's not broken.
>>>>>>>>>>>>>>
>>>>>>>>>>>>>
>>>>>>>>>>>>> What you need to do is use the information mdrun provided
>>>>>>>>>>>>> you to
>>>>>>>>>>>>> diagnose what's going on. Apparently atoms 193657 193660 are
>>>>>>>>>>>>> separated
>>>>>>>>>>>>> by 31 nm. What are your box vectors? Where are these atoms in
>>>>>>>>>>>>> the
>>>>>>>>>>>>> system? Then you'll have your answer. The only reason I can
>>>>>>>>>>>>> think
>>>>>>>>>>>>> of for
>>>>>>>>>>>>> such extreme distances is a periodicity issue.
>>>>>>>>>>>>>
>>>>>>>>>>>>> -Justin
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>
>>
>>
>

-- 
Fabio Affinito, PhD
SuperComputing Applications and Innovation Department
CINECA - via Magnanelli, 6/3, 40033 Casalecchio di Reno (Bologna) - ITALY
Tel: +39 051 6171794  Fax: +39 051 6132198