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

From: Charo del Genio <>
Date: Wed, 19 May 2021 18:25:51 +0100

On 19/05/2021 04:41, Liao wrote:
> Dear Amber community,
> A feature I found in my simulation these few days was that when I applied Cartesian restraints (ntr=1) to part of my protein (e.g. terminal residues or Zinc finger so that the zinc ion doesn’t fly out), the flexibility of this protein itself, and also movement of another bound protein, apparently increases. When I change the restraints into internal distance restraints (nmropt=1) for the same region, such increased conformation search speed is gone. Things slow down apparently.
> Why is this happening, my rough guess is that this maybe has to do with the Langevin thermostat; when using absolute restraints, the random forces from the thermostat is not dissipated into translational and rotational motion, therefore it turns into internal energy to make the protein flex around much more.
> I set gamma_ln=0.01 in opc water.
> Is this phenomenon commonly seen, or somewhat seen, by others also? And if so, was the reason what I had speculated? Very curious.

Dear Liao,
        I can't really say what causes the increased motion you see with Cartesian restraints, but I have a few comments nonetheless.

First, if you have a zinc finger, or indeed any other metal coordination centre, I suggest you try and parametrize it, rather than just putting a zinc ion in its vicinity and hoping it stays there.
For this, you can use MCPB together with GAMESS (or GAUSSIAN), which does a pretty good job.

Second, what do you mean exactly by internal distance restraints? Do you mean emap restraints or distance/angle restraints? Either way, it's not surprising that you don't see much conformational
changes in this case, since you are quite literally restraining the local or mid-range geometry of your molecule.

Finally, I feel that gamma_ln=0.01 may be a bit low for explicit-solvent simulations. Of course, the optimal value of many parameters depends on the system, but in your case I would probably use a
gamma_ln of the order of 2.0.



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 Wed May 19 2021 - 10:30:03 PDT
Custom Search