Re: [AMBER] Conformation search speed increase with Cartesian restraints

From: Charo del Genio <>
Date: Thu, 20 May 2021 12:40:15 +0100

On 19/05/2021 22:02, Liao wrote:
> Dear Charo,
> Thanks for your answers and comments,
> Talking about gamma_ln values, for the past few months I am still trying to gradually digest the effect of its different values on the speed conformation search. The reason why I used 0.01 is because based on this paper (, they reported the fastest conformation search using the smallest friction coefficient, although that is for implicit solvent. Why would explicit solvent be different? The water molecules may slow things down for all values of gamma_ln, but it shouldn't change the trend from what I think?
> Going back to an old thread from 2015 (, the paper suggested by Dave (Langevin dynamics of peptides: The frictional dependence of isomerization rates of N-actylananyl-N'-methylamide ) I had carefuly read, for their small system of 12 atoms they tested, it did not have solvent in it as I understood. They concluded gamma_ln=2 gives the fastest conformation change, based on equilibrium population and time to equilibrium. Although when they change the analysis criteria for calculation of the reaction rate to simple counting of barrier crossing events, one that considers recoils (product moving back to reactant?), it is 0.01 that is fastest. On the marco-level, kinetics definitely should be change in reactant/product population over unit time, but on the microscopic level, maybe there's a broader definition?
> As for the internal distance restraints I was talking about, what I mean is that I either use Cartesian restraints to hold the Zinc and its neighboring residues fixed; or I apply a distance restraint to keep the Zinc atom at a fixed distance from its neighboring sulfurs. So the same groups are being held in the two scenarios, just in a different way.

Dear Liao,
        I think I understand your questions. The behaviour of folding in explicit solvent with low values of gamma_ln has also been investigated in this paper If
you check their results, they report that too low gamma_ln results in unnatural dynamics of the system. I believe this is a good reason not to lower the viscosity too much, certainly in explicit
solvent. Note that, also in implicit solvent, I have once experienced unphysical behaviour of the protein I was modelling with gamma_ln=0.01. The problem disappeared as soon as I increased gamma_ln.
Incidentally, that protein also included a zinc finger.

Finally, if you are using restraints to keep the zinc and/or the coordinating atoms in place, you should absolutely try and parametrize the metal coordination centre, which will free you from such needs.



Dr. Charo I. del Genio
Senior Lecturer in Statistical Physics
Applied Mathematics Research Centre (AMRC)
Design Hub
Coventry University Technology Park
Coventry CV1 5FB
AMBER mailing list
Received on Thu May 20 2021 - 05:00:02 PDT
Custom Search