Re: [AMBER] ReCpHMD Optimal values of ntcnstph, ntrelax, nstlim and numexchg

From: Jason Swails <>
Date: Mon, 5 Dec 2016 10:56:45 -0500

On Mon, Dec 5, 2016 at 7:46 AM, Fabrício Bracht <> wrote:

> Hi. The first part of the question is regarding ntcnstph and ntrelax
> variables. Since the ntrelax value controls the solvent relaxation period
> after a state change, should the ntcnstph times the number of titratable
> residues be larger than the ntrelax value?

​Note that in implicit solvent, *every* titratable residue attempts to
change its protonation state (independently in random order) every ntcnstph
steps. Note that this is very different than implicit solvent, where only
a single residue is attempted every ntcnstph steps. The reason for this is
primarily one of computational performance, but the arguments in the manual
explaining how to pick ntcnstph based on the number of residues you're
titrating does not apply in explicit solvent.

ntrelax should be chosen to ensure reasonable solvent relaxation (the
longer the better) and reasonable computational performance (the shorter
the better). My original paper walks through this choice.

> Also I was wondering if there is some sort of consensus for good (wouldn't
> even ask for optimal) values of nstlim and numexchg for replica exchange
> constant ph simulations in explicit solvent.

​I like using 100 for both ntcnstph and nstlim (numexchg should be chosen
based on how long of a simulation you want to run). Replica exchange
attempts are *virtually* cost-free, although synchronization imposes more
overhead in explicit CpHMD simulations than any other simulation (due to
replicas having different numbers of relaxation steps to execute and the
ones with fewer having to wait for the replicas with more relaxation to

> I mean, how many protonation state exchanges have to be made within an
> nstlim window?

​They are independent sampling processes, so there are no restrictions.

I would recommend consulting the paper introducing the method, as it
addresses several of these questions directly.


Jason M. Swails
AMBER mailing list
Received on Mon Dec 05 2016 - 08:00:03 PST
Custom Search