On Wed, Sep 16, 2015 at 10:44 AM, Arjun Sharma <arjunsharma83.gmail.com>
wrote:
> Dear Jason,
>
> The reference energy values calculated for amino acids in CpHMD i.e
> Aspartate
>
> refene2 = _ReferenceEnergy(igb1=21.4298008, igb2=26.8894581,
> igb5=26.5980488,
> igb7=23.4181107, igb8=26.3448911)
> refene2.solvent_energies(igb2=33.2613028)
> refene2.dielc2_energies(igb2=12.676908, igb5=13.084913)
> refene2.dielc2.solvent_energies()
> refene2.set_pKa(4.0, deprotonated=False)
>
>
> was carboxyl proton on syn- or -anti position ? Do we need to calculate
> for both positions and average it out as there is only one value for
> particular IGB model ?
What I did was put it on the syn- position of a single oxygen and run TI.
(changeProtState in ParmEd can set the protonation states for you based on
the contents of the titratable residue python module) -- this TI was done
on AS4 with, for instance, state 1 chosen. Then what I did was run a
titration using this reference energy and computed the actual pKa from the
pH-REMD simulation. It did not come out fully correct due to entropic
contributions of the syn- and anti- populations (that are not degenerate,
so you cannot do a simple average). Using the difference between the
target pKa and the *actual* pKa, I was able to "fix" the reference energy
to its correct value. [1]
> Also can you please explain in little more detail the
> refene2.solvent_energies . I remember you mentioned before that same IGB
> model is used but with explicit dynamics. I’m not really clear on this part.
>
It is the slightly-adjusted reference energies for running in explicit
solvent (the difference being the generated conformational ensembles
between implicit and explicit solvent). These differences will in general
be very small (and I use the same approach to optimize the solvent
reference energies as I did for the original ones I mentioned above). The
reason AS4 and GL4 reference energies are different is because I had to
modify the GB radii of the oxygens for explicit solvent to get reasonable
results. So it's a completely different reference energy calculation than
in implicit. If you look at the definitions for the other
(non-carboxylate) residues, though, you will find that the solvent energies
and non-solvent energies are very close.
> Lastly, I see you also calculated the reference energies at int dielec 2.
> Why is that?
>
It took me a long time to diagnose the GB radius issue as the culprit for
poor performance of the explicit solvent model. In the process of trying
to improve the model, I tried *many* things, one of which was using an
internal dielectric constant of 2 (which obviously necessitated
recalculating the reference energies). It did not work (as it was not the
culprit), but the reference energies are still correct if you want to run
with a dielectric constant of 2, so I just left them in there. I wouldn't
personally suggest using it.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Sep 16 2015 - 08:30:10 PDT
