[gmx-users] Re: Use pull code to restrain the COM

[Justin Lemkul]

On 8/9/13 12:48 PM, Bin Liu wrote:
> Hi Justin,
>
> Thank you for your quick reply.
>
> In Setting A, grompp reports
>
> Pull group  natoms  pbc atom          distance at start
> reference at t=0
>         0           0             0
>         1           146        23185                -0.000 -0.000 -3.268
>    6.668  6.693  6.492
>
> The values of reference at t=0 are from the .mdp file. They are the COM of
> pull group 1 calculated by g_traj -ox -com. But I can not understand the
> values of distance at start. Where does -3.268 come from? If pull group 0,
> the reference group is set to be 6.668  6.693  6.492, and grompp can
> calculate the COM of pull group 1, the distance should be 0.0 0.0 0.0
> instead of -0.000 -0.000 -3.268.
>

True.  That seems buggy, but given that what you're doing is generally not 
recommended, probably no one has really investigated it.

> In Setting B, grompp reports
>
> Pull group  natoms  pbc atom        distance at start
>   reference at t=0
>         0           0             0
>         1           146        23185             -0.000 -0.000 -3.268
>   0.000  0.000 -3.268
>
> It seems Setting B doesn't agree with Setting A. Then pull-start=yes
> doesn't have the effect as I guessed.
>

It does, but it's probably impacted by incorrect behavior.

> My situation is a bit complicated and challenging. I am simulating the
> oligomerization of lipo-peptides inside a lipid bilayer. There are several
> lipo-peptides in the system. Initially I placed them in the water, i.e.,
> outside the bilayer. Then I found it takes too long (at the scale of
> microseconds even seconds) for them to spontaneously insert into the
> bilayer and form an oligomer.  Then I artificially placed the lipopeptides
> into the bilayer, and put them together to create an artificial oligomer. I
> want to see if they can stay there. But obviously the artificial insertion
> and oligomerization is unfavourable from the viewpoint of energy. So the
> lipopeptides have a strong tendency to leave the bilayer. Now I want to
> restrain the COM of each lipopeptide (at least in z direction) for some
> time, but give them conformational and orientational degrees of freedom as
> much as possible, to give rise to the possibility that the artificial
> oligomer can find a favourable conformation and stay there after the
> removal of restraints. Unfortunately GROMACS can not restrain the COMs of
> several groups simultaneously. Then in one short simulation (say 100ps), I
> can only restrain the COM of one lipopeptide and freeze the others (in z
> direction). The whole process proceeds as:
>
> MD (restrain COM of peptide A, freeze BCD) -> CG EM -> MD (restrain COM of
> B, freeze ACD) -> CG EM
> -> MD (restrain COM of C, freeze ABD) -> CG EM -> MD (restrain COM of D,
> freeze ABC) -> Next cyle -> ....
>
>>From my current experience, the EM step is necessary as in MD steps the
> freezed lipopeptides developed a large amount of strain. If the strain can
> not get released, the MD runs will explode in several hundreds of
> picoseconds (GROMACS reports failures of LINCS and eventually segmentation
> fault, an obvious indication that the system has exploded.) And pull-k1= 50
> is indeed very small. The restrained lipopeptides can displace in z
> direction by a quite large amount in only hundreds of picoseconds.
>
> Could you suggest a different approach, or some suggestions to refine this
> process? Thank you very much.
>

If you're experiencing such bad forces, I doubt you'll ever get a COM restraint 
sufficient to counteract the interactions with the membrane.  Moreover, even if 
you restrain the COM to allow the protein to freely interact with the lipids, 
the magnitude of the forces will probably yield severe artifacts.  You need a 
better method of inserting the peptides.  InflateGRO or g_membed are viable 
options.  We used the former in some recent work studying peptide 
oligomerization in membranes (http://pubs.acs.org/doi/abs/10.1021/bi400562x), 
and others have applied different methods for different systems.

-Justin

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

Justin A. Lemkul, Ph.D.
Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441

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