Re: FEP question

From: Victor Ovchinnikov (ovchinnv_at_MIT.EDU)
Date: Thu Sep 06 2007 - 11:29:23 CDT


Piggybacking onto Matthew's question: if it were possible to keep the
nonbonded interactions intact within the ligand while turning the
ligand/protein interaction off/on, would this be a better/legitimate
solution to the problem?


PS Matthew, FYI, there are theoretical papers discussing FES, e.g.
Boresch et al, J Phys Chem B 2003 107 p9535
Boresch & Karplus J Phys Chem A 1999 103 p103

On Thu, 2007-09-06 at 10:28 -0400, Jerome Henin wrote:
> Matthew,
> I think your ideas are essentially the right things to do.
> In principle, you can use a custom Tcl script that enforces soft
> restraints to, as you said, keep the two rings roughly in the same
> place.
> When dealing with isolated molecules, this is not a problem at all,
> since at the end-points (lambda=0 or 1), one of the molecules is
> virtually non-existent, so tethering it to the other does not bias the
> energies. Only points at 0 < lambda < 1 are affected, which just means
> that the free energy is computed along a different pathway - a
> better-behaved pathway, actually.
> Now, the issue with your DNA residues is that even at the end-points,
> the ghost residue still has bonded terms, so tethering it may induce a
> bias.
> Still, I see two possible courses of action:
> 1) add such restraints coupling the two residues, but in a way that
> does not add a significant conformational bias - this may be possible
> if "native" conformational fluctuations are small.
> 2) as you also suggest, add torsional restraints (same kind of custom
> Tcl script) to the ghost residue until its nonbonded interactions are
> strong enough. The issue there is that in theory, it would require an
> FEP step to determine the FE difference between the restrained residue
> and the unrestrained one... this may be done by growing the restraints
> using the "conformational free energy" module of NAMD, but it is still
> adding a nontrivial step to a nontrivial calculation.
> You see, there is no cheap way out, but I hope it helps nonetheless.
> Best,
> Jerome
> On 9/6/07, Matthew WIlce <> wrote:
> > Dear NAMDers,
> > I have a couple of questions regarding FEP.
> >
> > I am using FEP and mutating a Cytosine into a Thymine, ADE and GUA. I
> > have set up the dual topologies so that it is the purine or
> > pyrimidine that is mutating while the rest of residue in not
> > changing. I have noticed that at the beginning of the FEP run the
> > purine or pyrimidine that is starting to grow moves around too much
> > and sometimes the ring flips ~180 degrees. When the ring flips 180
> > degrees it prevents a sensible outcome from the FEP. This also occurs
> > for the Cytosine ring towards the end of the FEP. Once lambda has
> > become greater than about 0.01 and the growing ring starts to
> > interact with the rest of the molecules it behaves much better.
> >
> > For a purine to purine mutation eg ADE to GUA I could of course
> > modify the dual topology so that the all of the common atoms are
> > maintained throughout the FEP, and the same goes for the pyrimidines.
> > But I cannot see how a work around when the rings are quite different
> > eg. pyrimidine vs purine.
> >
> > Is there a way to sensibly restrain/constrain the movement of the
> > growing component of the FEP so that it cannot 'float' around until
> > lambda1 passes say 0.01, and also apply similar restraints to the
> > shrinking component when lambda 1 is greater then ~0.99.
> >
> > Has anyone done this? What impact could this have on the free energies?
> >
> > I have thought of writing a tcl script that constrains the atoms at
> > the beginning and end of the FEP depending on lambda ?? but I am not
> > sure how. I want the growing and shrinking rings of the purine or
> > pyrimidine to stay roughly in the same place relative to each other
> > until their interactions with the rest of the molecule may cause them
> > to move if necessary and not have them moving because they are not
> > interacting.
> >
> > I have tried using the namd parameters as suggested in the most
> > recent FEP tutorial as well as varying them but I have not been able
> > to resolve this. Even using a very low number of steps in each FEP
> > window (as low as 1000) does not stop the ring from sometimes flipping.
> >
> > Thanks in advance,
> > Matthew
> >
> >

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:12 CST