[gmx-users] Free energy profiles with Pull code : dimer/dimer interactions

[Justin Lemkul]

On 11/2/16 12:24 PM, CROUZY Serge 119222 wrote:
> Dear gromacs users
>
> In previous mail on the list Justin argued about free energy calculations with Pull code that:
> "For general
> protein-ligand complexes, it is *NOT* appropriate to assume
> one-dimensional pulling or applying position restraints to the
> protein.  Our paper (from which the tutorial was derived) describes
> the somewhat unique case we were dealing with. "
>
>  In my case, I'm calculating the profile for dimer dissociation in a
>  tetramer of proteins pulling on 1 dimer in direction X (because the
> tetramer is oriented in a such a way that X is the principal axis)
> while the other dimer is harmonically restrained (BB atoms only)
> I don't see how this could be done differently; although I'm hoping that
>  the free energy will result from the sole force between dimer/dimer
>  atoms: the restraint on one dimer should not affect the result .. As
>  long as they do not prevent the fluctuations (breathing) necessary for an almost reversible extraction of
> one dimer  from the tetramer. Is there a paper describing the possible effect of
> these restraints on the profile ?  (I suppose if one dimer is
>  completely fixed, the energy barrier for extraction will increase
>  dramatically because its residues won't be allowed to adjust while the
>  partner dimer is pulled out)
>

A Google search of "effects of restraints on potential of mean force 
calculations" returns numerous papers with methods detailing how to de-bias the 
calculations from the use of harmonic restraints.  In many cases, the effects 
are indeed minimal, or at least within the error of the PMF itself.

My commentary was intended to alert people of the potential errors of blindly 
following a tutorial and assuming that approach is universally applicable. 
Despite the fact that I warn against it in the tutorial, link my paper that 
describes why we did what we did, and comment on this forum regularly trying to 
dispel any false notions, I often see people propagating my tutorial files into 
other systems to which they are not applicable. A similar approach may, of 
course, be appropriate depending on the context.  In general, protein-protein 
binding is not necessarily one-dimensional (usually it's not) but a pathway 
defined for a higher-order complex may require additional assumptions about path 
that are up to you to defend in interpreting your results.

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================