[gmx-users] Coupling groups - Thermostat

Mark Abraham Mark.Abraham at anu.edu.au
Tue Jan 10 23:59:33 CET 2012

On 11/01/2012 6:52 AM, Justin A. Lemkul wrote:
> Steven Neumann wrote:
>> On Tue, Jan 10, 2012 at 7:07 PM, Justin A. Lemkul <jalemkul at vt.edu 
>> <mailto:jalemkul at vt.edu>> wrote:
>>     Steven Neumann wrote:
>>         On Tue, Jan 10, 2012 at 6:55 PM, Steven Neumann
>> <s.neumann08 at gmail.com <mailto:s.neumann08 at gmail.com>
>> <mailto:s.neumann08 at gmail.com <mailto:s.neumann08 at gmail.com>>__>
>>         wrote:
>>            On Tue, Jan 10, 2012 at 6:22 PM, Justin A. Lemkul
>> <jalemkul at vt.edu <mailto:jalemkul at vt.edu>
>> <mailto:jalemkul at vt.edu <mailto:jalemkul at vt.edu>>> wrote:
>>                Steven Neumann wrote:
>>                    Dear Gmx Users,
>>                     My system includes: ions, water, two tubes made of
>>         carbon
>>                    atoms, protein.
>>                    I would like to run NVT (and then NPT) with position
>>                    restarined dynamics of my protein and tubes.
>>                    I am wondering whether this approach is good (two
>>         coupling
>>                    groups: Protein_Tubes and Water_and_ions??
>>                     My thermostat in mdp file:
>>                     Temperature coupling is on
>>                    tcoupl = V-rescale ;
>>                    tc_grps = Protein_Tubes Water_and_ions ; two coupling
>>         groups
>>                    tau_t = 0.1 0.1 ; time constant
>>                    ref_t = 298 298 ; reference temperature
>>                    Please, let me know whether this apporach is ok. How
>>         can I
>>                    set tc_grps when I want to add ligand?
>>                I don't know a definitive answer here, so I'll throw 
>> out some
>>                ideas and hopefully stimulate some discussion.  I create
>>         tc_grps
>>                based on species whose dynamics are intimately 
>> linked.  For
>>                solvent, that includes water and ions.  Are your 
>> protein and
>>                tube physically associated?
>>            They are not physically associated but I put my protein as
>>         close as
>>            possible to the tube and I want to run position restrained
>>         dynamics
>>            of my tube and first 4 residues of my protein (stimulating
>>         attached
>>            protein to my tube).
>>         Will you suggest attaching my protein directly to my tube in
>>         this case?
>>     I'm assuming by "attaching" you mean coupling in the same tc_grp?  I
>>     wouldn't. This is a complex case (and again, I don't know a true
>>     answer here) - your system has the potential to be highly dynamic.
>>      Say the protein and tube bind, in which case they would (in theory)
>>     be coupled together.  Say they never bind, and then if you couple
>>     them together they shouldn't be.  You don't know a priori which way
>>     it will go.
>> No. I mean physically attached. That is why my first 4 resiudes are 
>> closed to the tube and position restrained. The best would be to 
>> attach it physically by sharing one atom. No clue how. My tube is a 
>> representation of the rest of the protein assembly (I am interested 
>> in the influence of charged "residues" represented by ions and non 
>> charged by carbon atoms within my tube - position restrained dynamics 
>> the tube) on my protein. What is more there is another tube above my 
>> protein (not attached) and I am interested also on the influence of 
>> those "residues" of the tube on my protein conformation. In future I 
>> want to do Umrella Sampling pulling my tube above to see free energy 
>> difference.
> It would have been better to state all of this up front ;)
> If you are trying to create a single entity representing the protein 
> and the tube, then yes, they should be a single tc_grp, and the best 
> approach is to create a merged [moleculetype] definition with an 
> actual bond between the shared atoms.  Doing so is not trivial, and I 
> have no real quick way to suggest doing that, other than recreating a 
> topology from pdb2gmx and perhaps making use of specbond.dat.  I'm 
> still not 100% clear on where everything is located.
> The other tube (not attached) and solvent/ions/ligands should likely 
> be treated as separate groups for the purposes of temperature 
> coupling.  But again, as stated, this is a very complicated system and 
> the best methodology for simulating it is likely not defined terribly 
> well.

One alternative is to pay attention to the advice at the end of section 
3.4.8 of the manual and ref cited there - that separate T-coupling 
groups can be worse than the problems they purport to fix.


More information about the gromacs.org_gmx-users mailing list