Re: convergence of abf

From: Eric H. Lee (ericlee_at_ks.uiuc.edu)
Date: Thu Sep 18 2008 - 05:19:49 CDT

Jorgen -

I believe the .abf file you refer to is the default ".dat" file - its
the file that contains the information for the free energy, average
force, and number of samples. If you did not specify a name in your
configuration file, then it defaults to "abf.dat." The .dist file
however, is an optional file that you must explicitly specify in your
namd configuration file under "distfile." Like the "history" file (I
usually call it .hist), it is an optional output file from ABF
simulations using namd. The history file is useful for evaluating the
convergence of your system. More details can be found here:

http://www.ks.uiuc.edu/Research/namd/2.6/ug/node35.html

- Eric

On Sep 18, 2008, at 5:04 AM, Jorgen Simonsen wrote:

> Hi Eric,
> thanks for your help - but I don't have any .dist and .dat files. I
> have the .abf which contains the results and the number of samples.
> I have the restart files *vel, *xsc etc.
>
>
>
>
> On Fri, Sep 12, 2008 at 11:19 PM, Eric H. Lee <ericlee_at_ks.uiuc.edu>
> wrote:
> If you are referring with the 15/17 as Angstrom extension along your
> reaction coordinate, then one way to solve a sampling issue would be
> to select a very narrow sampling window that focuses on that part of
> the extension. You can use as input the .dat and .dist files from
> the previous runs, thereby saving you the hassle of stitching the
> pieces of the PMF together.
>
> -Eric
>
> On Sep 12, 2008, at 10:03 AM, Jorgen Simonsen wrote:
>
>>
>>
>> Hi Eric Lee,
>>
>> Thanks for your reply - I have another peptide where the sampling
>> is not uniform. Would one than sample the areas which were not
>> sampled as much and than average against the main simulation - like
>> if you have made a 5 ns simulation and from 15 to 17 the sample
>> size is very small so I would than make a new simulation from 15 to
>> 17 and combine the results with the "old" simulation. But when you
>> combine should it be a weighted average - A = 10 for 10.000 and A =
>> 4 for 90.000 ?
>>
>> Thanks in advance
>>
>> Best regards
>>
>> J
>>
>>
>>
>> On Mon, Sep 8, 2008 at 9:45 PM, Eric H. Lee <ericlee_at_ks.uiuc.edu>
>> wrote:
>> At 5ns and 100,000 sample size, thats essentially your reaction
>> coordinate split into 50 sampling windows. If there are any
>> distinct barriers encountered along the way (essentially a sudden
>> rise in the PMF), you may wish to sample that event a little bit
>> longer. However the uniform distribution spread over your entire
>> sampling time is a pretty good indication of convergence.
>>
>> -Eric
>>
>> On Sep 8, 2008, at 9:16 AM, Jorgen Simonsen wrote:
>>
>>> Hi all,
>>>
>>> I have performed a ABF simulation where I am trying to stretch a
>>> peptide - the samples are more or less uniform distributed where
>>> the sample size is around 100.000 but how can one be sure that the
>>> energy has converged? My sampling time 5 ns which is a ilttle
>>> short but.....
>>>
>>> Any advice or comments appreciated
>>>
>>> thanks in advance
>>>
>>> Best
>>> J
>>
>> Eric H. Lee
>> Medical Scholars Program
>> Theoretical and Computational Biophysics Group, UIUC
>> ericlee_at_ks.uiuc.edu
>>
>>
>>
>
> Eric H. Lee
> Medical Scholars Program
> Theoretical and Computational Biophysics Group, UIUC
> ericlee_at_ks.uiuc.edu
>
>

Eric H. Lee
Medical Scholars Program
Theoretical and Computational Biophysics Group, UIUC
ericlee_at_ks.uiuc.edu

This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:49:50 CST