Re: [AMBER] Idea of equilibration in MD before production?

From: David A Case <>
Date: Mon, 18 Jun 2018 08:48:08 -0400

On Mon, Jun 18, 2018, senal dinuka wrote:
> I am writing here hoping for a clarification. I understand that doing
> minimization is to reduce bad contacts. And equilibration is mainly to give
> velocities given that system is per-heated to a certain temperature. So is
> it appropriate only to run few ns of equilibration so that the density is
> converged and move on to about 100-150ns of production phase?

Equilibration also allows the system to relax from its starting
configuration into the ensemble of configurations favored by the force
field. Depending on the size and floppiness of your system, and on the
quality of your starting configuration, this can take a long time. A
key parameter is the height of the free energy barriers that separate
various parts of conformational space.

How long to "equilibrate" is a complex function of the nature of your
system, how important it is that you are sampling at equilibrium, how
you are planning to use the results, and how much computing power you
have available.

I recognize that this is all rather vague, but the answer really depends
on your system. A small and fairly rigid molecule in water should reach
equilibrium in tens of nanoseconds or less. I have seen (folded) proteins
where the average structure was still drifting after several hundred
nanoseconds. But be aware the "equilibration" is not the only goal in
most simulations: in "production" (almost?) no biomolecular simulations
every done actually sample the full equilibrium configuration space.

So, one should try to design simualtions that can provide useful
information within the constraints of available computational power.
I like emphasize the goal of making it likely that one could learn
something new, that is, to get more information out of the simulation than
goes into it.


AMBER mailing list
Received on Mon Jun 18 2018 - 06:00:04 PDT
Custom Search