[gmx-users] Partial Charges PM3 ?

[Hugo Verli]

Dear Marc,

the adequacy in use semi-empirical charges for MD simulations depends on the
structure of the compound to be simulated. For example, compounds presenting
sulphur or phosphorus in its oxidized forms are not well described by
semi-empirical methods due to the absence of d orbitals in the calculations (do
not use MNDO/d method, since it does not present any improve in description of
organic molecules). Another point that is frequently raised when using atomic
charges from quantum methods in MM studies regards to the used atomic charge
scheme. Generaly charges derived from the molecular electrostatic potential are
better then Mulliken derived charges. However, properties based on
electrostatic potential are highly dependent on the molecular conformation (so
choose carefully the conformation to derive the charges). These two schemes of
atomic charges can be obtained with semi-empirical methods. An aditional method
is the Lowdin scheme, that while is much less dependent on molecular
conformation than the electrostatic potential derived charges, is highly
correlated with it (Becker et al, Carbohydr. Res. 340, 2005, 1499-1507).

If you are not intended to perform simulations with sulphate or phosphate
groups, semi-empirical charges should work fine. Try to use RM1 semi-empirical
method (Rocha et al, J. Comp. Chem. 27, 2006, 1101-1111,
http://www.rm1.sparkle.pro.br/), much better than PM3 and possibly than PM5,
but free. If not, is better to use ab initio derived charges (you can minimize
the molecule with 3-21G basis set and further perform a single point
calculation with 6-31G** basis set, for example).

Additionally, you should have in mind that the properties of a molecule in a MD
simulation with solvent and/or protein are not equally dependent on the atomic
charge. For example, we compared different basis sets on its abilities to
reproduce polysaccharide geometry in solution (Becker et al, Carbohydr. Res.
340, 2005, 1499-1507). We observed that the modification of the atomic charge
scheme does not interfere in the  molecular geometry, but it does modify the
interaction energy with the surrounding medium (solvent or protein). However,
it does not prevent you to observe relevant biological phenomena, as ligand
induced conformational changes (Verli & Guimarães, J. Mol. Graph. Mod. 24,
2005, 203-212). If you use the same protocol to obtain the atomic charges,
probably the error associated with the charges will be linearized.

I hope this helps.

Best regards,

Hugo Verli.

> Hey all,
>
>       I would appreciate an opinion on this. Is it acceptable to caluclate
> partial charges using a semi-empirical method (PM3) for topologies obtained
> frmo PRODRG? Is there any vital information that gets missed or skewed by
> doing this?  I have noticed that some QM/MM strategies will use
> semi-empirical methods for the QM side of the analysis and so was wondering
> if the same rationale can be applied to the caluclation of partial charges,
> or is this an apples to oranges type of comparison?
>
>   Regards,
>
>   Marc
>


Prof. Hugo Verli
Departamento de Produção de Matéria Prima
Faculdade de Farmácia - UFRGS
Av. Ipiranga 2752, Porto Alere, RS
CEP 90610-000, Brasil
tel.: +55 51 3316 7770
      +55 51 3316 5526

--------------------------------------------------------------------
Centro de Biotecnologia da Universidade Federal do Rio Grande do Sul