On Wed, May 11, 2016 at 7:28 AM, Mary Varughese <maryvj1985.gmail.com>
wrote:
> sir
>
> I mean it needs long time to converge and has to run separate trajectories.
> is it possible to finalize the calculation with 20 or 30 ns?
> i tried a 10 ns separate trajectory method as well as 15ns single
> trajectory. it fails.
> The ligand do bind throughout the simulation time but delta PB and TdeltaS
> as seen in MMPBSA calculation varies significantly especially TdeltaS
> varies from negative to positive values. (average of such varying values is
> not reliable isnt?)
>
> The ligand is very much flexible with CH3-CH3-CH3 chain. The ligand in
> individual structure is different from bound structure.
>
​Just because single trajectory MM/PBSA calculations have less variability
in the binding results does not mean that it is more converged than the
multiple trajectory approach. Single trajectory analyses are *perfectly*
correlated -- the bound and unbound conformations are exactly identical.
As a result, the variability of the binding energies is, by definition, at
a minimum. But this correlation does not exist in real life -- bound and
unbound conformations are never identical.
Naturally once you move to a multiple trajectory approach, the binding
energies will be a lot more variable and you won't get the illusion of a
converged simulation. There are other approaches to compute free energies
of binding based on MD, but none will be cheaper or easier than simply
running longer simulations for MM/PBSA. You will encounter the same
sampling problem with any method you choose. If you're encountering
accuracy problems, you can try a more rigorous method (like thermodynamic
integration or umbrella sampling/steered MD), but those will be a *lot*
slower.
HTH,
Jason
--
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed May 11 2016 - 05:30:07 PDT
