On Thu, May 12, 2016 at 12:26 AM, Mary Varughese <maryvj1985.gmail.com>
wrote:
> Sir,
>
> I am attaching herewith the output i got from single trajectory approach:
> Here deltaG(= -15.91-(-22.43=6.52 a +ve value). As you could see especially
> entropy varies much. Taking an average value does'nt seem to be ideal.
>
I don't know what you mean by "doesn't seem to be ideal". Taking an
average is what you should be doing according to the theory behind MM/PBSA
-- the snapshots are already sampled according to a Boltzmann distribution,
so the Boltzmann weighting is built into the average. That said, the
approximation used to compute entropy makes a lot of assumptions that may
not hold up, and its reliance on rigorous minimization for normal mode
calculations (which is difficult to do for large biomolecules) will likely
result in potentially noisy data. After all, the most sensitive eigenmodes
of the Hessian are the low-frequency modes, which also contribute the most
to the entropy. That's why the standard deviations of the -TS term are so
much larger than the other terms for the single-trajectory binding results.
>
> And i need to clear one thing about multiple trajectory.
> step 1: run three long separate trajectories say a 50 ns each.
> step 2:................
> doubt : usually in single trajectory approach if we create say 3
> snapshots;
>
> rec1: lig1: cmplx1==> deltaPB1 and TdeltaS1
> rec2: lig2: cmplx2==> deltaPB2 and TdeltaS2
> rec3: lig3: cmplx3==> deltaPB3 and TdeltaS3
>
> PB total= average(PB1,PB2,PB3); TdeltaS total= average(TS1,TS2,TS3);There
> is a correspondence between rec, lig and cmplx
>
> In multiple trajectories there is nothing, I may never get a
> correspondence.
> I must be missing something in the way doing multiple trajectory approach.
> It should be like Find the energy of rec, lig and cmplx for each
> trajectory(for the full length of simulation) separately and then find the
> difference at the end only. Isnt?
> I really dont understand the logic in there.
>
I recommend going to the literature describing the MM/PBSA protocol (start
from the MMPBSA.py reference and mine references from there and from the
Amber manual). Then, as the manual suggests, make sure you can run MM/PBSA
analyses *by hand* (i.e., without using either mm_pbsa.pl or MMPBSA.py) to
make sure you understand what is being done at each step. Many of your
questions would be easily answered by an understanding of the underlying
theory behind MM/PBSA, and such an understanding will undoubtedly help
you apply this methodology effectively now and in the future.
HTH,
Jason
--
Jason M. Swails
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu May 12 2016 - 11:00:03 PDT
