[gmx-users] trjconv, pbc and breaking of multiple peptides

[Justin A. Lemkul]

ms wrote:
> On 10/11/10 21:34, Justin A. Lemkul wrote:
>>
>>
>> ms wrote:
>>> On 09/11/10 21:36, Justin A. Lemkul wrote:
>>>
>>>> There are numerous -pbc options with trjconv; have you tried others? I
>>>> have never had luck with -pbc nojump actually working, and -pbc atom
>>>> provides no guarantee that molecules will be properly reconstructed.
>>>> Using -pbc mol -center is often a much better approach, in my
>>>> experience.
>>>
>>> I spoke too soon. -pbc mol -center doesn't indeed break molecules, but
>>> it doesn't seem to always work correctly -I evidently continue to have
>>> frames where the system is obviously not centered correctly in the
>>> box, so even if I have a compact conglomerate of my molecules, the
>>> calculated gyration radius is artificially large (In fact, I
>>> discovered the issue by looking at the frames with large gyration
>>> radiuses).
>>>
>>> I am sure it doesn't work correctly because if I visualize the system
>>> in VMD, enabling periodic copies I see the correct (compact) structure
>>> at the corners between boxes. - I can provide pictures if needed.
>>>
>>> Any idea on how to proceed? I am quite lost. I wouldn't like to
>>> rewrite g_gyrate by myself :)
>>>
>>
>> I presume you're centering on "Protein" in this case?
> 
> Yes, sorry for not making it clear.
> 
>> You can create a
>> custom index group (perhaps towards the center of one of your peptides)
>> and center about it. The -center function will locate the center of mass
>> of the chosen group at the center of the box, so, in a simple case (i.e.
>> a dimer) you would have two peptides at the corners, but the center of
>> mass still located at the box center. Choosing one residue of one of the
>> monomers has, in my experience, solved this issue when centering about
>> Protein has not worked.
> 
> I don't get it.
> 
> First, let me see if I understand it correctly. From your explanation 
> (and intuition), it seems that the issue is that "center of mass" in a 
> periodic environment is ambiguous -there are always (at least) two 
> alternate configurations that have the center of mass at the center of 
> the box any time, and -center doesn't necessarily choose the most 
> compact one (i.e. the one where the box center is closest on average to 
> atoms). So, for example, in a 1-D periodic environment, this:
> 
> ... O|O      O|O ...
> 
> is equivalent to this
> 
> |   OO   |   OO  ...
> 
> the center of mass between the O's being always at the center of the box 
> in both cases.
> 

Correct.

> But I don't see how choosing one residue could help -I mean, the 
> residues wander practically everywhere in the box during the simulation 
> and none of them is privileged ; I don't see why centering only around 
> one would be better. It could help if they were only two, but we talk of 
> eight objects here (and possibly many more in the future)...
> 
> ...now I see it: You mean that, since I force to center around one of 
> the objects, and if on average it should belong to the cluster, the 
> whole thing should be more often than not within the box?
> 

That's it.  Take, for example, peptides A, B, and C.  Peptide A is a nucleation 
point for B and C to attach.  Conceivably, trjconv could choose a "center" that 
involves C on one "side" of the box and A+B on the other.  The whole protein is 
still "centered" by its criteria.

If, physically, A is in the core of the aggregate, then I could choose some 
residue in peptide A and choose to center about it.  There is no other alignment 
that trjconv can assign, other than to place that chosen residue of peptide A 
right in the middle of your unit cell.  It can't be broken any other way.

-Justin

> I'm going to try; meanwhile let me know if I got it right (and just in 
> case, I wonder how much work is to patch trjconv to "choose" the "right" 
> center of mass -I guess it's similar to what -pbc cluster does? I never 
> fully understood that option)
> 
> m.
> 

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================