Hi,
> a)
> "restraints are divided by lambda" which is not in agreement with my just
> commented experience from which is clear that the restraints are divided
> by (1-lambda).
yes, that seems to be a typo, divided by (1-lambda) would be more correct.
> b)
> The scaling of the restraints (that reported in output files) is present
> probably just in case of usage of softcore potentials not in the case when
> the softcore potentials are not used
> and just simply linear mixing is used instead. Am I right ?
that is right. The two steps envisioned are
I) activate restraints without softcore and linear mixing of restraint
strength
II) keep restraints intact (by scaling) and switch of the ligand (using
softcore)
> -------------------------------------------------------
> Scaling up weight of 20.000 by 2.000
> New weight 40.000
> --------------------------------------------------------
>
> is really just informative and the final restraint applied to the ligand
> will be still original value of 20 in this case. Am I right ?
Correct, this is what the scaling message tries to say (maybe it could be
a bit better worded, though...)
> If this is true, the first step of the decoupling of the restrained
> ligand:
>
> dA*
> (RL)s -----> (R)s***(L)v
>
> (i.e. gradually turn on restraints while el./vdw interactions are
> simultaneously gradually turned off using softcore potentials)
>
>
> of the whole process
>
>
> dA* dAr
> (RL)s -----> (R)s***(L)v----->(R)s + (L)v (1)
>
>
> might be in principle realised in one Amber TI step by providing for each
> lambda different RST file (let say RST_lambda) where I use
> different values of force constants in each RST_lambda (i.e. smoothly
> growing with increasing lambda in my particular case).
in principle yes, in practice no, due to two problems: First, I do not
know how sampling will be affected by simultaneously turning on restraints
and turning off interactions, this would need to be carefully tested. The
second point is technical. Since the ligand is present only in V0,
restraints can be defined in V0 only. In this case they can either be not
counted in dvdl (thats what dvdl_norest does) or they count as restraints
being removed, since they are not present in V1 anymore. What you would
want, however is to count restraints as appearing while the ligand is
disappearing, something that Amber is not set up to do. You could subtract
the restraint contribution times two from the resulting dvdl by hand, but
would need to carefully check that this really covers all energy
contributions.
> (dAa*, dAb*). It seems to me (if I understood everything well) that you
> agree here with me just for the substep "dAa*" and in case of
> "dAb*-step" (gradual vanishing of vdw/el interactions of the fully
> restrained ligand) you suggest dvdl_norest=1.
dvdl_norest should only be switched on during the second, ligand
decoupling step. Turning it on in the first step defeats the purpose of
that step which is to calculate the dG-contribution of adding the
restraints. Also, since the first step does not turn on ifsc, you should
not be able to set dvdl_norest=1.
There is no problem with such long mails, Marek and I try to answer them
if I have time and believe that I understand the tricky points myself. I
still hope that running it over the list will hopefully inform a few
people more than it confuses ;-)
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 Tue May 08 2012 - 10:00:03 PDT
