Re: overlapping windows in ABF

From: Olya Kravchenko (
Date: Tue Jul 24 2018 - 18:37:44 CDT

Hi Jerome,

thank you, this is helpful. The gradients look somewhat different in one of
the overlapping regions, which is also less sampled, I may have to alter my

I am also trying to find out whether NAMD considers output from one file
(say, the first listed in the merger input) or from both files for the
overlapping region. I don't see this mentioned in any of the tutorials.

Thanks a lot,


On Mon, Jul 23, 2018 at 5:17 PM Jérôme Hénin <> wrote:

> Hi Olga,
> for a quick diagnosis, you could plot the free energy gradients from all
> windows on the same graph. Overlapping regions should have similar values
> of the gradient (keeping in mind that the gradient is always "noisier" than
> the free energy itself, so there can be some variation). Any inconsistency
> indicates that the two windows haven't sampled the same values of other
> important degrees of freedom. See the literature on ABF for more info on
> this, eg. this:
> Jerome
> On Mon, 23 Jul 2018 at 16:55, Olya Kravchenko <> wrote:
>> Dear all,
>> In my current ABF simulations two out of eight windows overlap slightly
>> with neighboring windows, i.e. instead of 52-56, 56-60 etc. they are
>> defined as 52-56.5, 56-60.5 (by accident).
>> I tried both merging files as is as well as merging them after removing
>> the overlapping portion of the data (i.e. cutting off data for 56.-56.5 and
>> 60.-60.5) and resulting pmf plot looks different after modification.
>> My question is, how does namd treat such overlaps? Should I concatenate
>> overlapping windows manually without cutting off the data for overlapping
>> 0.5A?
>> Thanks in advance!
>> Olga

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:18 CST