Re: alchemical FEP question

From: Sébastien Légaré (Sebastien.Legare_at_rsvs.ulaval.ca)
Date: Sun Dec 21 2008 - 20:03:07 CST

Thank you for the quick reply Christophe,

I want to use single topology to do a FEP pKa calculation. I do not expect a
great vdW contribution because the acid hydrogen is found within the vdW
radius of its connected oxygen according to CHARMM27 parameters.
Electostatics then becomes the major contribution and I think I can increase
λ quadratically as you and Dr. Pohorille propose in chapter 2 of "Free Energy
Calculations". I am not sure I can still do this if the appearing hydrogen is
allowed to move away from the vanishing oxygen.

Unfortunately, I can not afford to save all configurations of my trajectory. I
usualy save one every 1000 steps. I prefer to compute U1-U0 during the
simulation. As I started discussing it in the thread "Single topology FEP"
answered by Jérôme, I thought of emulating single topology by using strong
harmonic restraints between corresponding atoms of each states. I would then
divide by half the mass of all duplicated atoms. This would also require to
modify bonded force constants to keep vibration frequencies even with reduced
masses. Every single topology atom would then be replaced by two atoms of
half its mass strongly bound to each other. Does that make sense?

Honestly, I was not aware of the PMF-bond correction. Thank you for the
references. If I understand well, this correction could be needed in dual
topology too. The disappearing part could have shrinking bond lengths.

Thanks again

Sébastien

On December 21, 2008 02:27:57 pm Chris Chipot wrote:
> Hi Sébastien,
>
>
> now that a soft-core potential is available in the upcoming
> 2.7 version of NAMD to avoid end-point singularities, thanks
> to Floris Buelens and Chris Harrison's efforts, implementation
> of the single-topology paradigm would make little sense. Why?
> Under most circumstances, electrostatic decoupling would be
> necessary, zeroing off the charges and the Lennard-Jones
> parameters separately, with a potential of mean force correction
> to account for shrinking bonds. This translates to running two
> independent simulations, when the dual-topology requires only
> one. Yet, if you insist upon going for the single-topology route,
> your script ought to do the following: (i) drive NAMD to generate
> the trajectory representative of the reference state 0, (ii) alter
> the PSF by updating the charges according to q(λ) = λ q1 + (1-λ) q0
> (first leg of the alchemical transformation), (iii) alter the
> parameter file by scaling the Lennard-Jones parameters according
> to Rij*(λ) = λ Rij*1 + (1-λ) Rij*0 and εij(λ) = λ εij1 + (1-λ) εij0
> (second leg of the alchemical transformation), (iv) loop over the
> configurations representative of state 0 and evaluate the potential
> energy for state 1, (v) determine the ensemble average
> <exp [-β(U1-U0)]>0. In principle, a PMF-bond correction (see
> Pearlman and Kollman, as well as Hermans, back in 1991 for further
> detail) ought to be measured as well. I may still have a script that
> does the trick, but I would have to dig it out from archives.
>
>
> Chris
>
> seleg25_at_rsvs.ulaval.ca a écrit :
> > Hi Christophe and Namd users,
> >
> > I am interested in that script that would make Namd pretend it do single
> > topology (without fixing atoms). I am working on one but if there is
> > already one somewhere I would like to see it.
> > Are there any plans of implementing single topology in Namd?
> >
> > Thank you
> >
> >
> > Sébastien Légaré
> > Ph. D. Student
> > Laval University, Quebec
> >
> > Selon Christophe.Chipot_at_edam.uhp-nancy.fr:
> >> Hi Nandhini,
> >>
> >> what you are proposing to do is a single-topology transformation,
> >> wherein only
> >> the electrostatic component of your potential energy function is
> >> altered.
> >>
> >> In Namd, the paradigm chosen for alchemical transformations is
> >> dual-topology,
> >> where the topologies of both the reference and the target states
> >> coexist.
> >>
> >> Though you could cook up a quick-and-dirty script that would do the
> >> trick and
> >> pretend that Namd handles a single topology, I would advocate that you
> >> stick to the dual-topology paradigm, branching out your molecule into
> >> two topologies
> >> where the charge distribution starts to differ (consider the charge
> >> distribution
> >> of the initial and the final states).
> >>
> >> Chris
> >>
> >> Selon Nandhini Sokkalingam <at5629_at_wayne.edu>:
> >>> Hello all,
> >>> I am trying to mutate an aromatic compound to another one. The charges
> >>> on
> >>
> >> the
> >>
> >>> atomic sites for the initial and the target molecule are different
> >>> while
> >>
> >> some
> >>
> >>> of the
> >>> sites have similar vdw parameters. can i fix the positions of the
> >>
> >> specific
> >>
> >>> atoms with
> >>> similar vdw parameters to account for only electrostatic perturbation
> >>> and
> >>
> >> not
> >>
> >>> for
> >>> the non-bonded perturbation like in the charging of the sodium atom
> >>> alchemical
> >>> FEP tutorial.
> >>>
> >>> Thanks
> >>> Nandhini

-- 
Sébastien Légaré
Ph. D. Student
Laval University
Quebec, Canada, G1V 0A6
418-656-2131 #11577

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 05:21:36 CST