Re: [AMBER] mmGBSA bondi/mbondi2 and force field questions

From: Hai Nguyen <nhai.qn.gmail.com>
Date: Mon, 3 Jun 2013 18:23:56 -0400

On Mon, Jun 3, 2013 at 12:42 PM, <psu4.uic.edu> wrote:

> Dear Amber,
>
>
> After reading a couple papers, mailing list and user manuals, we are not
> 100% sure if the following choice for protein-ligand (small molecule)
> mmGBSA is justifiable.
>
> igb
>
> <*Preferable* bondi radii> (also doable)
>
> Protein Force Field
>
> Ligand Force Field
>
> 1
>
> *mbondi in mmGBSA*
>
> ff12SB
>
> GAFF
>
> 2
>
> *mbondi2 in mmGBSA* (bondi)
>
> ff12SB
>
> GAFF
>
> 5
>
> *mbondi2 in mmGBSA* (bondi)
>
> ff12SB
>
> GAFF
>
> 7
>
> *mbondi2 in mmGBSA *(bondi will cause instability in native peptide REMD)
>
> ff12SB
>
> GAFF
>
> 8
>
> *mbondi2 in mmGBSA** (*mbondi3 in MD/REMD)
>
> ff12SB
>
> GAFF
>
> In the previous mailing list, Hai and Jason mentioned that we should set
> mbondi2 for igb=8 in protein-small molecule mmGBSA so this part is answered
> already. However, in the paper of Hai et al. and Dill et al., they
> mentioned that mbondi2 might be the better choice for igb=7 for
> protein/peptide only simulation. Would you consider it would be the same
> in protein-small molecule mmGBSA?
>

I believe that you took this information incorrectly. In original igb=7
paper (Mongan et al.), our igb=8 paper (below), and Dill et al. showed that
igb7 is not a wise choice for either protein/peptide and nucleic acid
simulations. Dill et al. suggested to use the combination of GB5 + ff96 for
protein/peptide simulation. This is a kind of error cancellation between
GBOBC and the force field. This error cancellation is not always
transferred from one to another system. Our igb8 papers showed that igb=8
should make significant improvement compared to GB5,7 models.

http://pubs.acs.org/doi/abs/10.1021/ct3010485

We recommend to use mbondi2 with igb8 rather than mbondi3 since mbondi2 was
used to fit igb8 parameters. So if people only care about energy
correlation between GB and PB method, mbondi2 should be their choice.
However, in simulation mbondi3 should be used to get rid of strong salt
bridge bias in GB model (detail in our igb paper).

We have not done any comparison for mmGBSA calculation between different GB
(5,7,8) models and PB method yet. But igb8 theoretically should have better
correlation to PB than GB5 and GB7 since the energies for protein
structures calculated from igb8 are closer to PB energies than the one from
GB5,7. For small molecules, there is trivial different among those GB
models.


>
> The other question is: Amber12 and AmberTools13 manuals clearly states
> that we should use ff12SB for igb=7&8 but didn't recommend force field on
> igb=1,2,5. Would you recommend to run igb=1,2,5 on ff12SB? Or ff99SB,
> which igb=1,2,5 is frequently test on? Thank you.
>
> To me, looking for an error cancellation between GB model and force field
is not a good choice. It makes no sense to use a really bias force field
(for helical structure) while we have the most updated one (ff12SB), for
example. And igb1,2,5 strongly bias helical structure too.

Let me know if my explanation is not clear to you.

Best

Hai Nguyen

   Cheers,
> Henry
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Jun 03 2013 - 15:30:02 PDT
Custom Search