Re: [AMBER] mmP(G)BSA correlation time

From: Adrian Roitberg <>
Date: Thu, 05 Sep 2013 15:38:12 -0400

Hi Henry

The exact choice of when to decide if things have decorrelated is never
set in stone, but people usually look at either first time when you
cross zero, or maybe +0.1
Since the autocorelation decays quickly, it does not matter much how you
choose it, within limits.

In your case, you are interpreting the plot correctly. It shows that
your decorrelation time is roughly 700 ps according to your definition,
but I think you can safely say it is ~ 300 ps.

If you have 6000 frames, this means 'roughly' that only one every
6000/tau_corr is truly independent. This makes NO difference for the
average of delta G, but it does for computing error bars (standard
error of the mean).

Two more things.
1) I am a bit surprised that your decorrelation time is this large, I
was expecting something much shorter
2) the vertical tics in your deccorrelation plot are very very weird.
There is mathematics of physical system that should show that !


On 9/5/13 3:13 PM, wrote:
> Hi Jason and other gurus,
> Really thanks for your help. We have modified the script and generate
> correlograms.
> So after reading several text
> books<>,
> web info<>
> and
> Amber related papers about (auto)correlation, we still have a question:
> how to interpret the correlogram and know the correlation time from the
> plot:
> The attached is an example of the mmGBSA dG autocorrelation plot
> generated from the python API. (We take one snapshot from each ps MD
> trajectory. So here are one 6000 ps trajectory and 6000 snapshots of
> mmGBSA calculation.) So our understanding is that during around 700 ps, as
> noted in the red lines, the autocorrelation diminishes to the zero level
> (the following lag from 700 ps to 6000ps fluctuates around 0 level ). So
> the correlation time of this mmGBSA calculation time would be around 700
> ps?
> Cheers,
> Henry
> On Fri, Aug 30, 2013 at 12:29 PM, Jason Swails <>wrote:
>> On Fri, Aug 30, 2013 at 11:30 AM, <> wrote:
>>> Hello all,
>>> Thanks Jan-Philip's solution. It works! So it seems like our
>>> trouble-shooting was in the wrong direction : /
>>> However, there is a new error message pops out:
>>> Loading and checking parameter files for compatibility...
>>> Loading and checking parameter files for compatibility...
>>> Traceback (most recent call last):
>>> File "./", line 21, in <module>
>>> total_mut = data.mutant['gb']['complex']['TOTAL'].copy()
>>> *AttributeError: 'mmpbsa_data' object has no attribute 'mutant'*
>> This example was meant primarily to demonstrate a potential use of the API.
>> Attributes and data keys are created only as they're needed. Therefore,
>> mmpbsa_data will not have a "mutant" attribute unless you actually run
>> alanine scanning (thereby generating 'mutant' data).
>> As general advice with examples of this sort -- analyze the code in the
>> example so that you understand what each line is doing (this sample script
>> just computes and plots the autocorrelation function of the total complex
>> energies -- not likely of much use for what you want to do). It would be
>> good to cross-reference properties of the 'data' object with how the
>> AmberTools 13 manual describes the mmpbsa_data data structure. It also
>> never hurts to be comfortable with using the programming language used in
>> the example. FWIW, I'm hardly alone in thinking that Python programming is
>> potentially the most useful
>> A 'quick fix' for this error of course is to delete all references to the
>> mutant data. However, while that will give you the satisfaction of
>> watching the script finish and a graph appear, it is unlikely to provide
>> you with useful data.
>> HTH,
>> Jason
>> --
>> Jason M. Swails
>> BioMaPS,
>> Rutgers University
>> Postdoctoral Researcher
>> _______________________________________________
>> AMBER mailing list
> _______________________________________________
> AMBER mailing list

                             Dr. Adrian E. Roitberg
Colonel Allan R. and Margaret G. Crow Term Professor.
Quantum Theory Project, Department of Chemistry
University of Florida
AMBER mailing list
Received on Thu Sep 05 2013 - 13:00:03 PDT
Custom Search