On Wed, Dec 2, 2015 at 7:41 AM, Fabian gmail <fabian.glaser.gmail.com>
wrote:
> Dear Hannes,
>
> I bother you again….
>
> According to the Kaus paper, it’s clear that only one minimization of the
> ligand it’s necessary to calculate the solvation free energy, but it’s not
> clear to me if each lambda value requires a heating and pressure
> equilibration, and then production, or only one initial heating + equil is
> necessary and then production continues from the last .rst7 file…. of the
> last lambda (as in a regular long MD).
>
In principle, you will always need to do an equilibration. You cannot
assume that the phase space of different lambda values are the same (in
fact, they are NOT the same), which means that a snapshot from one lambda
value will not be an equilibrium structure for another lambda value in
general. So you will need to perform some amount of MD in order to relax
the structure to its "new" lambda value and throw out the non-equilibrium
structures generated on the way to the actual free energy surface.
The more similar these lambda surfaces are (which you can usually guarantee
by taking small increments of lambda), the less re-equilibration time will
be needed. The surface similarity can be a lot worse if all you want to do
is avoid reminimizing and reheating (which means that even if you have to
re-do the equilibration when moving to an adjacent lambda value, you can
usually re-use the minimization/heating results from the previous step).
That said, heating and minimization are the cheapest parts of the
simulation (by far). So there's not really much harm in just redoing it if
you were concerned about 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 Dec 02 2015 - 06:00:08 PST
