[gmx-users] qmmm and "real-life" meaning of null lennard-jones parameters

[Javier Cerezo]

The system in you topology must be complete even for QMMM, so you'll 
need FF parameters also for the QM part.

Partition models for QM/MM work with the following energy decomposition:
E(total) = E(QM_system) + E(MM_system) + E(QM-MM interaction)

E(QM_system) and E(MM_system) are easy to define (computed at QM and MM 
levels respectively), but E(QM-MM interaction) is a bit more involved. 
One way to account for this interaction energy is through the ONIOM (and 
the like) partition scheme:
E(total) = E(QM_system at QM level) + E(whole_system at MM level) - 
E(QM_system at MM level)

So, you need to evaluate the whole system at MM level and that's why you 
need a force field for this part as well. Anyway, I think that for the 
QM_system tretaed at MM level only nonbonded interactions mater, since 
bonded interactions (that are not involved with the boundaries) are 
cancelled out when performing this subtraction: E(whole_system at MM level) 
- E(QM_system at MM level). For that reason, I think that bonded parameters 
that are not present in the boundaries, are not relevant and could be 
omitted in the topology.

Javier



El 08/06/12 01:47, Justin A. Lemkul escribió:
>
>
> On 6/7/12 7:28 PM, Edward Deira wrote:
>> Dear all,
>>
>> I'm currently starting to dwell deeper in MD, and I'm taking some 
>> time to
>> understand what's going on inside the gromacs "black-box".
>> In one of those dwellings, I came across an older post
>> [http://www.mail-archive.com/gmx-users@gromacs.org/msg42568.html] 
>> which reads:
>>
>> Question:
>> 4. In ffnonbonded.itp, why are both sigma and epsilon set to zero for HW
>> (opls_117)? This seems to imply that, as far as Lennard-Jones 
>> interactions are
>> concerned, the hydrogens on the waters don't exist. Or, in other 
>> words, in the
>> absence of charges, the hydrogens don't "feel" the hydrogens, the 
>> hydrogens
>> don't "feel" the oxygens, and the oxygens don't "feel" the hydrogens. 
>> In other
>> words, the hydrogens interact with the world only via electrostatic 
>> (Coulombic)
>> interactions. Is this a correct interpretation?Correct. Many force 
>> fields do this.
>> Answer:
>>
>> So, my question, if a question at all:
>>
>> Suppose I have a regular protein and put inside some metal atom that 
>> will
>> coordinate with some O and N atoms from the side chains. If the sigma 
>> and
>> epsilon for that metal are null, than the metal - sidechains 
>> interactions are
>> exclusively electrostatic. Does this make sense ? What are the 
>> implications of
>> this for the "coordination chemistry" of that "metal - sidechain 
>> complex" ?
>>
>> On the side: suppose I want some non parameterized metal atom, say W, 
>> for which
>> I will compute all the other parameters in the same/similar way 
>> described in the
>> force field papers, but for which no experimental data are available 
>> for me to
>> compare computable meaningful sigma and epsilon values. Can I just 
>> sigma and
>> epsilon to zero ? Or should I do qmmm to have W in the qm part ?
>>
>
> The fact that the LJ parameters for H are zero derives from its size.  
> The environment is more strongly influenced by the heavy atom to which 
> H is bonded.  In the case of a larger metal ion, I would seriously 
> doubt that setting LJ parameters to zero is valid.  It's quite 
> convenient, but in most force fields, all metal ions have some LJ 
> parameters.  Perhaps investigating how those parameters were derived 
> would be useful.  For what it's worth, I believe the origin of the 
> zero-LJ H parameters comes from this work:
>
> http://pubs.acs.org/doi/abs/10.1021/ja00824a004
>
>> Also, from the few tutorials and from the manual, I have the 
>> impression that
>> even for qmmm with gromacs and mopac i still need force field 
>> parameters for the
>> qm part, is this true ? Or i just need to include a qmmm section in 
>> all mdp
>> files, including the first ion adding and energy minimization steps ? 
>> Sorry for
>> the naivety in this, but i've only made "regular protein" MD so far.
>>
>
> I've never done any QM/MM, but my assumption would be that you have to 
> have some valid topology to start with.  Perhaps someone else can 
> comment on this methodological issue.
>
> -Justin
>

-- 
Javier CEREZO BASTIDA
PhD Student
Physical Chemistry
Universidad de Murcia
Murcia (Spain)
Tel: (+34)868887434