From: Jérôme Hénin (jerome.henin_at_ibpc.fr)
Date: Tue May 02 2017 - 05:07:29 CDT
On 2 May 2017 at 11:15, <yjcoshc_at_gmail.com> wrote:
> Hello,
>
> 1. At this stage, you can use on-the-fly eABF to replace classical ABF in
> any free-energy simulations, while egABF is also worth to try in high
> dimensional free-energy calculations. On-the-fly eABF have three free
> energy estimators: Naive, CZAR and zheng/yang's.
>
Yes, all of these are possible, but for a study of ion permeation, there
are linear coordinates for which plain ABF will work fine. I would stick to
that for simplicity. One thing that will improve convergence, if that turns
out to be difficult, is the multiple-walker scheme (explanations and
references are in the documentation).
2. I am also puzzled with the total force change, but I have seen an
"subtractAppliedForce" option in the user guide. It can recover the
behavior before the change. If you have problems you can turn it on.
Essentially, the new behavior ensures that you get the same results with
ABF that you'd get with metadynamics or brute-force sampling, which I think
is a sound default.
subtractAppliedForce only *partly* recovers the old behavior: it makes ABF
ignore other biases applied to the specific colvars involved in ABF, not
forces from other sources applied to the same atoms. The bottom line is, if
you run ABF under a complicated set of biases, you probably need to think
hard about what those other biases mean and how they should affect the ABF
result. If you only run ABF and no other biases, then that change in "total
force" behavior has absolutely no impact on your work.
Jerome
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:20:16 CST