Re: Setting alchElecLambdaStart 1.0 parameter for FEP calculations

From: Brian Radak (
Date: Mon Nov 12 2018 - 12:01:45 CST

Your interpretation should be correct. Unfortunately, I don't think you can
accomplish what you want via NAMD keywords. However, it's not clear to me
that there is an advantage to your proposed scheme over doing separate
annihilation/exnihilation runs for each molecule separately. The only
difference is that your scheme would have a an extra bonded and LJ
interacting molecule floating about. That might actually be a harder
endpoint to sample than the other molecule completely decoupled.

*What you want could be implemented: *If you look at the scheduling of the
different energy terms, each type has a special form of the general scaling
function (in C-like ternary syntax):

f(x; xs, xe) = (x <= xs ? 0.0 : (x >= xe ? 1.0 : (x - xs) / (xe - xs)))

where xs is the "start" value and xe is the "end" value. For electrostatics
xe is *hardcoded* as 1 and for van der Waals xs is *hardcoded* as 0. I
believe what you want is to have alchElecLambdaEnd != 1, but this is not
currently implemented. There might be some tricks with extra PME grids in
order to get the right comparison energies. We're working to simplify a lot
of this in 2.14, but that is probably quite a ways down the road.


On Fri, Nov 9, 2018 at 1:10 PM David Huggins <> wrote:

> Hi all,
> I hope to clarify something about the alchElecLambdaStart parameter for
> FEP calculations:
> My reading of the manual is that if I set alchElecLambdaStart 1.0, then
> electrostatic interactions are completely decoupled for both annihilated
> AND exnihilated particles at all values of lambda.
> The behaviour I would like is that electrostatic interactions are
> completely decoupled for one species (eg exnihilated) but completely
> coupled for the other (eg annihilated) at all values of lambda. Is there a
> way to accomplish that?
> Thanks,
> Dave

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2019 - 23:20:20 CST