[gmx-users] Sampling windows

[XAvier Periole]

On Jan 5, 2011, at 6:03 PM, Gavin Melaugh wrote:

> Hi Xavier
>
> Many thanks. Two more quick questions (both refering to barrier  
> region):
> Does the value of the force constant affect the height of the barrier
> providing that there is reasonable overlap?
No! The force constant used does not affect the PMF or the barrier.  
Neither
does the number of windows used. The only thing that can affect your
PMF is a change in the sampling you put in ...
> Obviously there is a
> compromise between the size of the constant and the number of windows.
True! And the objective being to get equilibrated distribution with  
reasonable
overlaps.
> If I increase the force constant and number of windows to obtain a  
> high
> density of sampling, does this give a more accurate curve?
No! The only thing you need is to sample everything. Actually putting a
strong force constant might have the side effect to trap your system and
then you do not sample the "free" degrees of freedom! With weak force
constant you may have to wait very long to obtain an equilibrated  
distribution.

XAvier.
>
> Gavin
>
> XAvier Periole wrote:
>>
>> On Jan 5, 2011, at 5:01 PM, Gavin Melaugh wrote:
>>
>>> Hi Xavier
>>>
>>> Many thanks for the reply. When I compare the two curves
>>> Curve 1) lack of overlap in barrier region; force constant = 1000
>>> Curve 2) with increased number of sampling windows with increased  
>>> force
>>> constant (note I have only increased the force constant of the  
>>> windows
>>> in the barrier region).
>>> Force constant =1000, Force constant in barrier region  = 5000
>>>
>>> The barrier increases due to better sampling in this region  
>>> (expected),
>>> but the rest of the curve in 2) is shifted upward compared to  
>>> curve 1).
>>> Why is this so when the only difference is the increased number of
>>> windows and force constant in one region?
>> the relative position of the regions on the left and right of your
>> "barrier"
>> region is completely depending on the barrier region!
>>
>> you may have overestimated the difference between bound/unbound  
>> states
>> because the energy barrier region was not sampled.
>>
>> Note also that cure one is simply wrong so comparing to it is  
>> dangerous.
>>
>> You have to be careful that you might still not have a converged
>> sampling:
>> you might have overlap of your histograms on the reaction  
>> coordinate but
>> other degrees of freedom are not fully sampled ... that might  
>> result in a
>> funny and wrong profile. You might want to look at your simulations  
>> for
>> funny things ...
>>
>>>
>>> Gavin
>>>
>>>
>>>
>>> XAvier Periole wrote:
>>>>
>>>> On Jan 5, 2011, at 4:08 PM, Gavin Melaugh wrote:
>>>>
>>>>> Dear all
>>>>>
>>>>> I have generated a PMF curve for introducing a guest molecule  
>>>>> into a
>>>>> host molecule using the umbrella sampling method. The curve seems
>>>>> fine,
>>>>> but there are no overlaps between histograms in the barrier  
>>>>> region. To
>>>>> circumvent this I have added more windows in this region and to  
>>>>> ensure
>>>>> sampling have increased the force constant only in this region.  
>>>>> The
>>>>> windows in this problematic region are now not equally spaced.  
>>>>> Does
>>>>> this
>>>>> matter for g_WHAM?
>>>> Nope it does not matter.
>>>>> Could someone please reply to me on this matter as it
>>>>> has been a sticking point for me for quite sometime?
>>>>>
>>>>> Many thanks
>>>>> Gavin
>>>>> -- 
>>>>> gmx-users mailing list    gmx-users at gromacs.org
>>>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>>>> Please search the archive at
>>>>> http://www.gromacs.org/Support/Mailing_Lists/Search before  
>>>>> posting!
>>>>> Please don't post (un)subscribe requests to the list. Use the
>>>>> www interface or send it to gmx-users-request at gromacs.org.
>>>>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>>>
>>>
>>> -- 
>>> gmx-users mailing list    gmx-users at gromacs.org
>>> http://lists.gromacs.org/mailman/listinfo/gmx-users
>>> Please search the archive at
>>> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
>>> Please don't post (un)subscribe requests to the list. Use the
>>> www interface or send it to gmx-users-request at gromacs.org.
>>> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>>
>
> -- 
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/Search 
>  before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists