Re: [AMBER] TI electrostatic free energy]

From: Franz Waibl <>
Date: Tue, 1 Dec 2020 16:24:44 +0100

Dear Dr. Case,

This might be a very helpful comment, even though I don't yet fully
understand it...

It was my understanding that the internal energy contributions of the
"vanishing" state (TI region 1) would automatically be kept regardless
of lambda, since this is also done in the 1-step transform using
Softcore potentials, and since it seems to be a reasonable thing to do.

However, your comment suggests that this is not the case, and that I
need to run a separate TI calculation in the gas phase. Is this correct?
Furthermore, I now realized that the coordinates of TI region 1 and 2
are exactly equal, so there is never any sampling of the decoupled
state, which could explain why Amber behaves different here.

If this is correct so far, could you please clarify the following:

* How does Amber decide whether a TI region will be "decoupled" or
removed, i.e., whether the internal energy of that region contributes to
dV/dlambda? I found a notice in the Section of the Manual about Softcore
potentials which states that atoms are "decoupled", but is this only
true when using Softcore potentials?

* If I need to account for the gas phase ensemble myself, how should
this be done? Can i just subtract the according energy contributions
from the gti_output table, or do I need gas-phase TI calculations? (I
expect conformational sampling of the gas phase to be rather irrelevant
in my case, since all of my molecules are rather rigid.) In the latter
case, is it important to use the same periodic box, or is it better to
run non-periodic simulations?

Thanks a lot for your help!

Best regards,

Franz Waibl

Am 01.12.2020 um 15:08 schrieb David A Case:
> ...quick comment in case it is helpful - it was not clear to me from the
> original post what cycle was being used - and if, perhaps, Franz might only
> have been talking about one leg of the cycle and not subtracting of the gas
> phase, leading to an artificial internal electrostatic energy component that
> could be quite large.
> Dy
> On 11/30/2020 1:10 PM, David A Case wrote:
>> On Mon, Nov 30, 2020, Franz Waibl wrote:
>>> I am trying to compute the free energy of hydration of a set of small
>>> molecules at different salt concentrations, using TI.
>>> The problem is that I am getting very large free energies (e.g, 130
>>> kcal/mol for caffeine, favoring lambda=0).
>> Is there any chance that your caffeine molecule has a net charge? It
>> would also be helpful to plot <dV/dlambda> vs. lambda: it that
>> approximately a straight line? Is that value at lambda = 0.5 about 65?
>> I'm really guessing here, but this looks like results for de-charging an
>> ion. (Use the "summary" command in parmed to look for the charge
>> model.)
>> ....dac
> _______________________________________________
> AMBER mailing list

AMBER mailing list
Received on Tue Dec 01 2020 - 07:30:02 PST
Custom Search