Re: [AMBER] How to use the ntb options in amber-cuda-version

From: David Cerutti <>
Date: Tue, 1 Sep 2020 03:38:48 -0400

As stated on page 360 of the Amber20 manual, ntb = 1 stipulates constant
volume conditions, ntb = 2 constant pressure. If ntb == 2, you must have
some sort of barostat working ("barostat" keyword). There is one more
wrinkle here: if you want isotropic box size rescaling (which is
appropriate for proteins in water), specify ntp = 1. Otherwise specify ntp
= 2 for anisotropic rescaling in systems like crystal lattices. There is
even semi-isotropic rescaling (equal proportions in X and Y, independent
proportion in Z) for membrane simulations, where you have a system that is
reasonably isotropic in two dimensions (the lipid bilayer) and a third
dimension that should not be coupled to the other two.

You are correct about the equilibration protocol, and that the production
MD should occur at constant pressure in order to measure Gibbs free

You should do your initial 100-200ps of constant pressure MD in the CPU
code, unless you are confident you have added enough water to maintain the
volume you started with (which would probably also take a program like
AddToBox or CHARMM-GUI rather than the tleap water box tiling procedure).
I would recommend at least 10ns in NVT, with positional restraints on key
atoms, before you move on to the NPT step. The NPT can still have some
positional restraints, but make sure they are light (0.1 kcal/mol-A2) by
the time you start this. Start the restraints off at 1-5 kcal/mol (or
more), then ramp 'em down to the 0.1 range to start NPT. (Note that you can
use the CUDA code for the NVT part, then swap to the CPU code briefly, then
back to the CUDA code.) I would advise CPU code for the initial energy
minimization, too: the GPU code does some sophisticated things with its
precision format in order to make energy minimization possible, but just
understand that we had to go to some lengths in order to make it reasonably
safe. Safer still to just let the CPU work for five minutes on that stage.

Good luck!

On Tue, Sep 1, 2020 at 2:33 AM <> wrote:

> Hello, Amber users!
> Recently, I want to empoly Amber program packages to simulate a
> protein in water system in order to guide my wet experiments. I plan to use
> the cuda-version of Amber18 to run. But I had mentioned some tips from
> amber18 manual that was "Avoid using the NPT ensemble (ntb=2) when it is
> not required; if needed make use of the Monte Carlo barostat (barostat=2).
> Performance will generally be NVE>NVT>NPT (NVT~NPT for barostat=2)." In my
> knowledges, the protein-salt-water system was should be considered to
> simulate under the npt ensemble for final MD production. So, I'm confused
> whether to chose the option ntb=1 or ntb=2. I known that the
> protein-salt-water simualtion should be equilibrated by NVT and NPT
> ensemble before MD production. So, How should I process the simulation
> with amber18 cuda version. Please help me. Any suggestion is welcome.
> Best wishes!
> Ning Wang
> _______________________________________________
> AMBER mailing list
AMBER mailing list
Received on Tue Sep 01 2020 - 01:00:03 PDT
Custom Search