On Wed, Sep 09, 2020, Debarati DasGupta wrote:
>
>Since last week I have been looking into literature to search for growth
>of molecules rather than disappearing molecules in AMBER18-TI simulation
>protocols.
Some background here:
1. This sort of discussion is not about formal correctness, but about
efficiency: how can one get good estimates of free energy changes without
wasting simulation time? (Related questions: how can one estimate the
statistical uncertainty in the results, and avoid systematic errors?)
2. There are (generally speaking) two different approaches to alchemical free
energy calculations, thermodynamic intergation (TI) and free energy
perturbation (FEP). Since the days of Tom Darden, and helped
along some by my preferences/biases, the Amber codes have concentrated on TI,
with only fairly recent excursions into FEP. For example, Chapter 23 of the
Amber Reference Manual ("Free Energies") has 13 pages of discussion of TI
(section 23.1) and only one tiny section (23.1.9) on FEP.
Why does this matter? In TI, there is no notion of "direction" of the change.
All simulations are equilibrium ones, carried out at a particular value of
lambda, and have no notion of whether one is growing or disappearing any
atoms. So nothing will happen if you reverse the masks, other than changing
the sign of the free energy change.
[Caveat: the "direction" can matter if you use the end point of one lambda
simulation as the starting point for the next one. But this means you are not
really approaching equilibrium at each lambda value.]
Even FEP calculations by Amber-ites, mostly carried out by post-processing,
tend to be direction-agnostic, using changes in both directions in
Delta-lambda in a balanced fashion. Others on the list may be able to chime
in about FEP schemes more akin to those dicsused by Prof. Jorgensen in his
article.
3. It is a really advanced topic, beyond my abilities and certainly not well
done via email, to discuss the "TI vs FEP" question in general.
...hope this helps....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Sep 10 2020 - 06:30:02 PDT
