RE: More FEP using separated topologies

From: Radak, Brian K (
Date: Wed Apr 26 2017 - 09:59:47 CDT

I was actually about to respond with that same suggestion, but reconsidered because it is in fact just a horrible kludge. However, that is probably the only way at present to accomplish what you want using NAMD internals.

The language for alchDecouple is kind of horrible - the default (off) is to remove internal non-bonded interactions within each group. This only changes the endpoint at which the group is to be completely uncoupled.

Another option is to modify your PSF via ParmEd (, which generally handles such topologies quite well, even though it was originally designed for AMBER topologies.


Brian Radak
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory

9700 South Cass Avenue, Bldg. 240
Argonne, IL 60439-4854
(630) 252-8643
From: [] on behalf of Jrme Hnin []
Sent: Wednesday, April 26, 2017 5:27 AM
To: David Huggins
Cc: Namd Mailing List
Subject: Re: namd-l: More FEP using separated topologies

Hi David,

sorry, I missed your previous reply.

Looking back, my advice doesn't seem that good anymore. I meant for you to just use FEP, but I see now that it will not yield the situation you are aiming for, since at either end-point of the transformation, one of the ligands will be decoupled.

The only way I can think of, which is more of an ugly hack, is to abuse FEP at lambda = 0.5, where each ligand is half-coupled. For that you'd need to disable soft-core LJ potentials and set the perturbation schedule to plain linear:

alchVdwShiftCoeff 0.
alchElecLambdaStart 0.
alchVdwLambdaEnd 1.

and then tweak the nonbonded parameters so that the ligand interactions are twice as large (ie. double the epsilons and partial charges). Finally you'd need to ensure intra-ligand nonbonded interactions are perturbed as well, by keeping alchDecouple to its default value, off.

I hope I'm not forgetting something here.


On 25 April 2017 at 21:10, David Huggins <<>> wrote:
Thanks Jerome,

Please can you clarify what you mean? Do you suggest increasing the
cutoff or using a later version of NAMD?

Best wishes,



Dr David J. Huggins
University of Cambridge
Department of Chemistry
Lensfield Rd
United Kingdom

Phone: +44 (0)1223 763854


On 28 May 2016 at 05:00, Jrme Hnin <<>> wrote:
> Dear Dave,
> I was reminded today that explicitly excluded atoms are expected to lie
> within the cutoff distance. Those exclusions are not required (anymore) in
> standard FEP runs because those atoms belong to separate partitions. That
> might be the way to go for your needs.
> Best,
> Jerome
> On 27 May 2016 at 22:21, David Huggins <<>> wrote:
>> Dear All,
>> I am trying to implement a NAMD version of the separated topologies
>> method from "Separated topologiesA method for relative binding free
>> energy calculations using orientational restraints".
>> I have used alchemify to switch off interactions between two ligands
>> in a protein binding site and the protocol requires me to run dynamics
>> where both ligands are fully coupled to the system (though not to each
>> other).
>> However, the resulting simulations always exit early with a "Low
>> global exclusion count". I have tried gentle heating and reduced
>> timesteps but neither works.
>> Does anyone have any experience of running simulations where two fully
>> coupled species are overlapping and have no mutual interactions?
>> Best wishes,
>> Dave
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>> Dr David J. Huggins
>> MRC New Investigator
>> Theory of Condensed Matter Group
>> Cavendish Laboratory
>> JJ Thomson Avenue
>> Cambridge, CB3 0HE
>> United Kingdom
>> Phone: +44 (0)1223 764164
>> Fax: +44 (0)1223 337356
>> Email:<>
>> Web:>
>> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:13 CST