Re: ABF subdividing the reaction coordinate

From: Aron Broom (broomsday_at_gmail.com)
Date: Wed Jul 10 2013 - 17:44:57 CDT

Well, subdividing the problem into smaller segments has many benefits.
Some are:

1) you can parallelize the problem. So now you can run multiple shorter
simulations. Since in most cases computers are only so fast but you may
have access to many of them, its ideal to be able to have many independent
jobs to run, rather than one huge one.

2) If there is a particular problem at some reaction coordinate, say for
instance the system has some really slow degrees of freedom as it crosses
some energy barrier. You can spend extra computational time on that window
only, rather than having to run the whole thing for longer.

I admit I don't 100% understand your first paragraph. I think what makes
sense, and is what you are suggesting aswell is:

1) run SMD to "roughly" explore the whole reaction coordinate. Let's say
the range is 0.0 to 1.0.
2) take snapshots from that SMD run at: 0.0, 0.1, ... 0.9, 1.0
3) use those snapshots as the input coorindates for ABF. Where, for
instance, the snapshot at 0.3, would be used as the starting coordinates
for an ABF window whose range along the reaction coordinate was set to be
0.2 to 0.4 (using colvar boundaries).

If something else is specifically suggested in that reference, it's
probably a better idea, but the above is at least sensible :)

~Aron

On Wed, Jul 10, 2013 at 6:15 PM, Pedro Swagger <pedrito.swagger_at_gmail.com>wrote:

> Yes, I have the complete SMD trajectory. I don't know if I am
> understanding well, but you mean after I dividing my SMD simulation into
> windows for ABF, I can take any configuration of my system inside each
> window for starting the ABF simulation for each respective window, and
> after merging all of them together I should similar results as I do the ABF
> from the begining without using windows. Am I right?
>
> And last question why do we use that approach of using windows from SMD?
> Why no to do ABF without windows?
>
> I got the approach for using windows from the tutorial entitled "Forcing
> substrates through channels" but I am not getting the real meaning of doing
> that.
>
> Sorry for many question.
>
> Thanks again.
>
>
> On Wed, Jul 10, 2013 at 6:01 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>
>> So you're taking the trajectory from SMD and using points along that as
>> the starting points for your ABF? That sounds like a reasonable way to
>> proceed. Assuming your ABF converges fully, the results should be the same
>> as using any other starting point for the windows (again assuming
>> convergence).
>>
>>
>> On Wed, Jul 10, 2013 at 5:45 PM, Pedro Swagger <pedrito.swagger_at_gmail.com
>> > wrote:
>>
>>> Thanks Aron,
>>>
>>> By the way, I run first a SMD simulation and I am dividing that
>>> simulation in windows.
>>>
>>> I have another question. Since I am taking the reference points of the
>>> system from a previous SMD for each starting windows, would the results be
>>> similar to if I run my system from the beginning using only ABF with
>>> windows (point of reference)?
>>>
>>> Thanks again
>>>
>>>
>>>
>>>
>>>
>>> On Wed, Jul 10, 2013 at 5:22 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>>>
>>>> If you are going to start a new simulation or "window" you can
>>>> reasonably take any snapshot from the previous window, but ideally you
>>>> would take one in which the reaction coordinate is closest to the new
>>>> window. Better yet, since you probably want reasonable overlap between
>>>> adjacent windows, you can take a snapshot from the previous window in which
>>>> the reaction coordinate is actually in the next windows range.
>>>>
>>>>
>>>> On Wed, Jul 10, 2013 at 5:00 PM, Pedro Swagger <
>>>> pedrito.swagger_at_gmail.com> wrote:
>>>>
>>>>> Dear everybody,
>>>>>
>>>>> I am trying to use ABF and I have some questions. It may be a garbage
>>>>> questions, but I am new to this. Sorry and apologize me.
>>>>>
>>>>> When subdividing the reaction coordinate in equally spaced window for
>>>>> your ABF simulation, the starting point for the simulation (in each new
>>>>> window) should be the begining configuration of the protein in that window?
>>>>> or it could be any configuration inside the window.
>>>>>
>>>>>
>>>>> Thanks
>>>>> --
>>>>> Pedro Swagger
>>>>>
>>>>> Saludos
>>>>>
>>>>> Departamento de Física,
>>>>> Facultad de Ciencias,
>>>>>
>>>>> Universidad de Chile
>>>>> Las Palmeras 3425, Ñuñoa, Santiago, Chile
>>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> Aron Broom M.Sc
>>>> PhD Student
>>>> Department of Chemistry
>>>> University of Waterloo
>>>>
>>>
>>>
>>>
>>> --
>>> Pedro
>>>
>>>
>>> Saludos
>>>
>>> Departamento de Física,
>>> Facultad de Ciencias,
>>>
>>> Universidad de Chile
>>> Las Palmeras 3425, Ñuñoa, Santiago, Chile
>>>
>>
>>
>>
>> --
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
>>
>
>
>
> --
> Pedro
>
> Saludos
>
> Departamento de Física,
> Facultad de Ciencias,
>
> Universidad de Chile
> Las Palmeras 3425, Ñuñoa, Santiago, Chile
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:23:27 CST