Re: [AMBER] Questions (and a small bug report) about SIRAH in Amber

From: Charo del Genio <>
Date: Wed, 26 Aug 2020 16:38:37 +0100

On 26/08/2020 16:27, Matias Machado wrote:
> Dear Charo,
> Some of this questions are answered in the SIRAH FAQs (check the web under menu "TUTORIALS", or in the package at sirah.amber/tutorial/SIRAH_FAQs.pdf)
> You can equilibrate in NVT or NPT... the reason for choosing the later rather than the former is due to the initial solvent density of the box... be aware that using small pre-equilibrated solvent boxes to solvate the whole system is a fast but not perfect method... indeed, owing to the algorithm to add water constituents and PBC conditions, several molecules are removed from both the surface of the solute and the image boundaries... that leads to a decrease in the expected water density for any computational box... that issue is magnified in case of coarse-grained system as removing one molecule implies removing a lot of water... hence, we equilibrate in NPT to fix that issue, otherwise you will see vacuum bubbles during equilibration, which may o may not affect your system at production...

Dear Matias,
        thank you, I'm aware of the FAQs. Anyway, note that I'm not talking about equilibration in NVT, but rather heating the system from 0K in NVT, as opposed to assigning initial random velocities from
the Maxwell-Boltzmann distribution at the right temperature.

> Regarding thermostats...
> The parameters for the Langevin thermostat are optimized for SIRAH and its integration time step, take into account that other "common" values are for all-atoms and not coarse-graned. We can't grant you the same results if you play with that...
> The same goes for V-rescale, I don't know which values are you using or why you prefer this thermostat over the other... We haven't tested it yet, because it's new in AMBER, but in any case, I would start by using the same values we use in GROMACS simulations... those are tau_t = 1.0 for equilibration and 2.0 for production...

For this specific system, I prefer V-rescale because part of one of the two proteins I'm simulating is intrinsically disordered, and the V-rescale thermostat seems to give more realistic results about
the structure of such regions than Langevin, which I would otherwise normally use. Anyway, I'm using tautp=1.0, so I guess I need to find the reason elsewhere.

> Regarding the softness of the equilibration protocol...
> That may certainly depend on the system... the step-wise protocol provided by us was proved to work in many systems, however it's not perfect and neither is SIRAH... so it isn't unlike to fail...
> Best,
> Matias

And of course also here it may be a matter of tweaking.
Anyway, thanks again for your answers.



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 Aug 26 2020 - 09:00:02 PDT
Custom Search