From: Niel Henriksen <>
Date: Thu, 19 Oct 2017 12:24:19 -0700

Hi Andreas,

I've tested both the MBAR approach, as well the TI approach, for APR
calculations. The results are usually very similar for systems with
sufficient sampling. My experience is that MBAR can be a little more
tolerant of sampling problems/gaps, and so it has the potential to be more
efficient if you want to use as few sampling windows as possible. I often
run both calculations and compare, since a large difference between the two
approaches can suggest a sampling problem.

In this paper <> I briefly
compar the two approaches, but mostly focus on the difference in the
estimated uncertainty when using the blocking method or statistical
inefficiency approach, which is a somewhat separate issue from the
calculated free energy.


On Thu, Oct 19, 2017 at 9:04 AM, Andreas Tosstorff <> wrote:

> Dear all,
> I was wondering if there was any literature that compares WHAM, DHAM,
> MBAR, BAR, with TI. My understanding of all these methods is very
> superficial. I am currently using TI as implemented in the APR script by
> the gilson group. But I was wondering if using other methods to get the
> free energy of binding would also make sense or even improve accuracy.
> I would really appreciate any suggestions.
> Best,
> Andreas
> _______________________________________________
> AMBER mailing list
AMBER mailing list
Received on Thu Oct 19 2017 - 12:30:03 PDT
Custom Search