Re: [AMBER] Question for to parameterize a nonstandard residue for constant pH MD

From: Jason Swails <>
Date: Wed, 27 Aug 2014 17:26:55 -0400

On Wed, Aug 27, 2014 at 3:30 PM, M Olivia Kim <> wrote:

> Hello,
> I was trying to parameterize a nonstandard residue for constant pH MD in
> Amber 14 and noticed that
> $AMBERHOME/AmberTools/src/etc/cpinutils/ script has been
> primarily changed from the corresponding in Amber 12 or older
> version.
> Specifically, in the old, I only needed to get the "relene"
> term, which I obtain from TI computation. I guess this term corresponds to
> the _ReferenceEnergy term in the script in Amber 14. However,
> in the, there are a couple additional terms that I need to
> define: solvent_energies and dielec2_energies.
> Would anyone be able to explain 1) what those terms mean; 2) how I can
> obtain the values for each term; and 3) how they affect the total reference
> energy? Thank you!

​1) solvent_energies is the reference energy for running CpHMD in explicit
solvent. It is usually quite close to the "normal" TI in GB (since the
protonation state moves are still done in GB). But since conformational
ensembles may be slightly different, so too are the reference energies.

​dielec2_energies are the reference energies if you use an internal
dielectric of 2. That was something I was trying when I couldn't get good
results with CpHMD in explicit solvent. As the problem turned out to be
the effective GB radii increased by the 4 dummy hydrogens on the
carboxylate residues, ​I'd not recommend the dielectric 2 be used by

2) For the dielec2_energies, you can do a TI setting intdiel=2. If you
aren't planning on using it though, there's no point. For the
solvent_energies, you will need to use the original GB reference energies,
run a titration (with pH-REMD for accelerated convergence), and then adjust
the reference energy so that the computed pKa matches the reference pKa.
 This can be done in a single step (i.e., you don't need to iterate to get
the correct reference energy until the pKa is properly predicted) -- I can
describe exactly how if you need me to.

In all cases, it's best to run a quick test titration (on GPUs you can do
hundreds of nanoseconds on a model compound in a few minutes) and then
adjust the reference energy to get the experimental pKa.

3) They are different reference energies -- if you tell to give
you a CPIN file that you plan to use in an explicit solvent environment, it
uses the solvent_energies instead. Likewise if you plan to use an internal
dielectric of 2. You only have to add the energies that you actually plan
to use, though.

Hope this helps,

Jason M. Swails
Rutgers University
Postdoctoral Researcher
AMBER mailing list
Received on Wed Aug 27 2014 - 14:30:02 PDT
Custom Search