[gmx-users] Re: How to set the sigma and epsilon for Cu2+ in OPLS-AA/L

[Dr. Vitaly Chaban]

On Tue, Apr 9, 2013 at 11:03 AM, fantasticqhl <fantasticqhl at gmail.com>wrote:

>  Dear Dr. Vitaly Chaban,
>
> Thanks very much again. I am sorry for the unclear, charge transfer was
> also taken into account for the complex, I did not mentioned in the last
> e-mail.
>
> What do you mean by finite T effect in MD? Kinetics?
>
>

I mean thermal motion. You have an optimal structure/energy at 0K in QM. In
MD you want to simulate at higher T, I guess. The optimal structures in
both cases may be very similar, but may be not so similar. I am just saying
that you should not expect ideal coincidence of energy vs. distance curves.



> For the reproduction of binding energy, I guess I know how to do it using
> QM method. Simply, I just need to do three single point calculations for
> complex,
> ligands and ion, respectively.
>


And correct for BSSE.



> For MM method, it is similar, however, I am not sure I get get the MM
> energy for just one ion.
>


This energy is zero within classical MD, since you do not consider
electrons and nucleus, as you do in QM.

Only one calculation is needed for MM. You define the charge groups, such
as "ion" and "ligand" and look at the interaction between them (g_energy).







>  Is my understanding right?
>
> Thanks for all your explanations and suggestions on this problems!
>
> All the best,
> Qinghua Liao
>
>
> On 04/09/2013 10:03 AM, Dr. Vitaly Chaban wrote:
>
>
>
>
> On Tue, Apr 9, 2013 at 9:39 AM, fantasticqhl <fantasticqhl at gmail.com>wrote:
>
>>  Dear Dr. Vitaly Chaban,
>>
>> Thanks very much for your patient and detailed suggestions on this
>> problem. Actually, I am doing what your suggested now.
>> I optimized the copper-ligand complex using QM method, and then did some
>> QM scannings to derive the bond and angle force constants.
>> Right now, I am doing the MM scanning using the same coordinates which
>> were used in the QM scanning. What we want is that the MM curves
>> can reproduce the QM curves.
>>
>
>
>  I think it is simply impossible in your case to reproduce the QM curves.
> You neglect charge transfer from copper to the ligand, resulting a chemical
> bond formation, you neglect finite T effect in your MD. If you want to
> remain in the framework of LJ+Coulomb, the best think you can get is
> reproduction of ion-ligand binding energy and more or less adequate
> distance ion-closest atom of the ligand
>
>
>
>
>>  But some of them agreed well, some of them did not. So I try to tune
>> the sigma of the liganded atoms, however,
>> it is a little complicated to tune many liganded atoms at the same time.
>> I am still trying to work it out.
>>
>>
>
>  Start from the sigma for "ion-closest atom of the ligand". All other
> atoms will adjust automatically, since they are connected all together
> within the ligand.
>
>  My personal viewpoint, which you may share or not, is not to do anything
> with sigmas of other atoms of the ligand. It is best for future portability
> to limit refinement to the ion only.
>
>
>
>
>>  It seems that you have much experience on such problems, could you
>> please give me some suggestions on tuning the sigmas of atoms again?
>> Thanks very much in advance!
>>
>>
>> All the best,
>> Qinghua Liao
>>
>>
>>
>> On 04/08/2013 03:51 PM, Dr. Vitaly Chaban wrote:
>>
>>
>>  On Mon, Apr 8, 2013 at 3:36 PM, fantasticqhl <fantasticqhl at gmail.com>wrote:
>>
>>>  Dear Dr. Vitaly Chaban,
>>>
>>> Thanks very much for your patient explanation. Yeah, you are right, that
>>> is what I want to know: how you tuned this parameter?
>>>
>>> Since then, if I want to set a new atom type and I know its vdw radius,
>>> so how should I set the sigma for it based on the vdw radius,
>>>
>>
>>
>>  You cannot set the sigma based ONLY on the VDW radius.
>>
>>
>>
>>>  which should be in agreement with OPLS-AA/L force filed? Could you
>>> give me some suggestions?
>>> I guess that I have to tune it by myself this time, right? Thanks in
>>> advance!
>>>
>>
>>
>>  I would do the following:
>>
>>  1) Optimize ion-ligand complex using ab initio. Write down binding
>> energy and optimal distance;
>> 2) Construct topology for classical MD using approximate sigma;
>> 3) Calculate energy and distance from classical MD;
>> 4) Compare them to distance and energy from ab initio;
>>  5) If you are not satisfied, adjust your sigma;
>> 6) Repeat classical MD until the difference between ion-ligand distance
>> in classical MD becomes reasonably similar to that in ab initio.
>>
>>  To preserve compatibility with OPLS, use the same level of theory in ab
>> initio, which they used when derived OPLS. Keep in mind that their original
>> level of theory is not so perfect...
>>
>>
>>  Dr. Vitaly Chaban
>>
>>
>>
>>>
>>> All the best,
>>> Qinghua Liao
>>>
>>>
>>>
>>>
>>>
>>
>
>