Hi,
> #1
> I would like to know how from the MD run in some fixed value of lambda (L)
> is possible
> to get <dV/dL>. I would naturally assume that two MD runs (first with L
> and the second with L + dL) has to
> be done from which <V(L)> and <V(L+dL)> are obtained and then <dV/dL>
> might be estimated as
> ( <V(L+dL)> - <V(L)> )/dL or as the ( <V(L+dL,t) - V(L,t)> )/dL where t
> is the simulation time.
the lambda-dependent potential can be written down, so you can work out
analytical derivatives and calculate them at a single lambda-value.
> #2
> I also did not understand fully (as it is not explained anywhere) why for
> the MD run with the given fixed L value the two simultaneous
> sander simulations has to be done for any TI calculation or it.s stage
> (one with the prmtop_V0 file and the second with prmtop_V1 file)
> and how the data obtained from these two simultaneous simulations are
> combined at the end.
> Could you please put some light on this ? (e.g. using as the example the
> study from your tutorial A9 for
> the explanation what is going on in each of that two MD simultaneous
> threads during the all three stages)
For implementation details, please see our 2007 paper (Steinbrecher,
Mobley, Case, JCP). If you want to know why the Amber run is conducted in
the way it is, the best way to start may be the code, as the
implementation is quite straightforward. Check out thermo_int.f, where the
actual force and energy mixing is done.
> I can only guess that this approach allow for simultaneous vdw
> coupling/decoupling of the different atoms
> like in our case in second stage when benzene H6 is decoupled and phenol
> O1,H6 atoms are coupled.
yes, that is true, you can simultaneously have atoms appear and disappear.
> However this should be also eventually possible to divide into 2 separate
> stages ((i)only H6 disappearing and then (ii) only O1,H6 appearing)
> with separate dG contributions which together should get the dG of the
> simultaneous H6 disappearing and O1,H6 appearing or not ?
You could do that if you wanted with the currect code. However, I do not
see why you would want to break it down into these two steps. Removing and
adding a groups at the same time seems a better way to do this for me,
since it doesn't introduce yet another non-physical state quite different
from the end states.
> Regarding to uncharging of benzene H6 atom stage or the last stage where
> O1,H6 phenol atoms are charged I really have no
> explanation for two simultaneous sander simulations during the given
> lambda run.
one prmtop has charges on the atoms the other hasn't. Each is represented
as its own groupsander process. This could have been done in a single
sander simulation as well, but neither the simulation speed nor the
ease-of-use would have benefited from such an option in my opinion.
> #3
> I also do not understand why do you have three columns of the graphs here
> ( http://ambermd.org/tutorials/advanced/tutorial9/analysis.html )
> labeled as Step X - Complex, Step X - Water, Step X - Complex-Water where
> X is going from 1 to 3.
Well, as the column labels say, one gives the values for the
transformation in complex, one in water and the last one the difference
between both.
> #4
> Do you think that TI might be possible to use also for the calculation of
> the absolute free energy of binding
> of the small lignad (L) to the surface of bigger receptor (R) for example
> using this scenario.
>
> STARTING SYSTEM
>
> L bound to R in given water-box.
>
> FINAL SYSTEM
>
> L and the R separated in the same water-box sufficiently.
>
> (Of course assuming the identical positions of all atoms except L in the
> initial files for STARTING and FINAL system)
>
> So I assume that during the whole TI procedure divided into relevant
> decoupling/coupling (El.,vdw.) stages
> L continuously "disappear" from it's original position and "appear" on
> it's final position
> sufficiently far from the R in the same water box.
>
> Might be this approach OK in your opinion ? If not what is wrong ? How to
> improve it if TI could be really useful here ?
While you could do things this way, there is extensive literature on doing
absolute dG-TI (cited e.g. in our papers). The principle is simple, but
there are numerous pitfalls and details that people have worked out, so
make sure you consult these before attempting to setup a simulation.
> BTW do you recommend in such case which I described above to use TI
> (however maybe with much more complicated protocol than I suggested)
> or is better and also more common to use here Umbrella sampling ?
Depending on what your reaction coordinate for the dissociation looks
like, US may be a better option than TI to obtain the binding energetics.
I can't say anything definite here, since it really is very system and
structure dependent.
Kind Regards,
Thomas
Dr. Thomas Steinbrecher
formerly at the
BioMaps Institute
Rutgers University
610 Taylor Rd.
Piscataway, NJ 08854
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Dec 05 2011 - 04:30:02 PST
