[gmx-users] OPLS parametrizaton

[Justin A. Lemkul]

Marcelo Silva wrote:
> Em 02-04-2011 0:33, Justin A. Lemkul escreveu:
>>
>>
>> Marcelo Silva wrote:
>>> Hi,
>>>
>>> I am trying to simulate mixtures of ethanol and trifluoroethanol and 
>>> I would like to use the same parameters used on OPLS parametrization.
>>>
>>> On the OPLS1996 paper the authors starts by stating that they've used 
>>> 267 molecules and "The box size varied from approximately 26 26  26 
>>> Å for methanol to 37  37  37 Å for cyclohexane". In my simulations 
>>> I am considering using 500 molecules (each box vector would be around 
>>> 37 Å).
>>>
>>> "In most cases, the intermolecular nonbonded interactions were 
>>> truncated at 11Å based on roughly the center-of-mass separations with 
>>> quadratic smoothing of the interaction energy to zero over the last 
>>> 0.5 Å (...) A standard correction was made for the Lennard-Jones 
>>> interactions neglected beyond the cutoff."
>>>
>>> Trying to interpret what is written I think the correct way to treat 
>>> cut-offs and energy corrections would be like this:
>>>
>>> ---------------------------------------
>>> Cut-offs
>>>
>>> rlist                    = 1.4
>>> coulombtype              = Reaction-field-zero
>>> rcoulomb                 = 1.1
>>> rcuolomb_shift      = 0.6
>>> vdwtype               = shift
>>> rvdw                     = 1.1
>>> rvdw_switch        = 0.6
>>>
>>> Dispersion Correction
>>>
>>> DispCorr             = EnerPress
>>>
>>> ---------------------------------------
>>>
>>> However I have some doubts because I'm using a box with a different 
>>> number of molecules/ different size (should I change the cutoffs to 
>>> 15Å like it was for cyclohexane?) and I'm not sure about using 
>>> Reaction-field-zero instead of the shift function for coulombic 
>>> interactions.
>>>
>>> Could you please help me confirm this parameters so I can run a 
>>> simulation in the same conditions as OPLS was parameterized (although 
>>> using MC and not MD).
>>>
>>
>> I don't think there is a perfectly clear answer to this.  OPLS was 
>> initially designed for MC with some special functional forms and is 
>> now applied to MD using all manner of settings.  Commonly, one sees 
>> rcoulomb=rvdw=rlist=1.0 or 1.4 in conjunction with PME.  There have 
>> been a variety of systematic force field comparisons in recent years.  
>> You should consult those to see what their specific recommendations 
>> are with respect to proper settings, although many of them focus on 
>> proteins.
>>
>> -Justin
>>
>>> Best regards,
>>>
>>> Marcelo
>>
> 
> Just to test it, I used this options (nstlist = 5 with timestep 1fs):
> 
> ; Cut-offs
> 
> rlist                    = 1.4
> coulombtype              = Switch
> rcoulomb                 = 1.1
> rcuolomb_shift      = 1.05
> vdwtype               = Switch
> rvdw                     = 1.1
> rvdw_switch        = 1.05
> 
> ; Dispersion Correction
> 
> DispCorr             = EnerPres
> 
> and compared to vdw plain cutoff at 1.1 and PME with cutoff 1.1 nm 
> (rlist = 1.1 nm also). The big difference in the output seems to be with 
> the Coulomb (SR) term:
> 
> -1.25490e+05 kj/mol with Switch
> -5.38895e+03  kj/mol   with PME
> 
> With switch I get a low density for ethanol (about 0.3 g/ml). What could 
> be the problem with the simulation using switch? By the way does gromacs 
> use a different switch function depending on the forcefield?
> 

The manual warns of artifacts that arise from switching functions.  So there is 
not a "problem" with switch, per se, rather an intrinsic limitation in accuracy. 
  I would imagine the low density arises from completely neglecting (via 
truncation) any long-range electrostatics terms that PME (and to some extent, 
other long-range methods) otherwise includes.

The switch function is not force field-dependent.

-Justin

> Best regards,
> 
> Marcelo
> 

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

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

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