[gmx-users] Umbrella sampling

chris.neale at utoronto.ca chris.neale at utoronto.ca
Thu Mar 31 17:39:20 CEST 2011


can you show us your mdp and the pmf and the histograms for the data  
that you put into wham? It's a lot easier to diagnose with the real  
data.

In the general case where umbrellas are spaced equally along your  
reaction coordinate, sampling overlap between umbrellas will always  
decrease anywhere the PMF is concave down, such as your barrier  
region. I would suggest adding windows at every 0.005 nm spanning the  
region that you consider to have a sampling problem (e.g. 0.76 to 0.83  
?) and use a much stronger force constant here (e.g. 2-3 times as  
strong as you used for umbrellas with 0.02 nm spacing). You have  
identified a problem, so my suggestion is to not bother fiddling with  
adding one extra replica but sample that region to death with strong  
force constants. I presume that this will not impact your overall  
efficiency too much if the reaction coordinate is long overall.

Finally, you will need to evaluate the convergence of your PMF overall  
and perhaps of this region in particular, especially if you want to  
know the dG to cross it or between two low energy states on either  
side of it.

Chris.

-- original message --


Sorry I am not sure that I follow. Will the window with r0 =0.80 giving
the distribution centred around 0.78nm not drive my free energy profile
up. If I remove this window prior to running g_wham the free energy goes
down. Should I increase the force constant so that the mean of the
window is 0.80nm (bearing in mind that this is near the barrier region).

Gavin
XAvier Periole wrote:
>
> You can present the data differently:
> you have two windows at 0.78 nm giving different distribution.
>
> That indicates these windows are not converged. Does not mean
> that the others (0.80 nm) are converged :))
>
> On Mar 31, 2011, at 12:20 PM, Gavin Melaugh wrote:
>
>> Hi Xavier
>>
>> Thanks for the reply. With respect to your answer of my first query.
>> What if you had two windows practically on top of each other, but one
>> was not supposed to be there. e.g A window with r0 of 0.80 nm and
>> centred at 0.78 nm and a window with r0 of 0.78 nm centred at 0.78nm.
>>
>> Gavin
>>
>> XAvier Periole wrote:
>>>
>>> On Mar 31, 2011, at 11:53 AM, Gavin Melaugh wrote:
>>>
>>>> Hi All
>>>>
>>>> I have generated several PMF curves for the one system using umbrella
>>>> sampling. In the first part of the curve (barrier region) I use a high
>>>> force constant with small intervals between the windows. The latter
>>>> part
>>>> of the curve I use a lower force constant with larger window spacing.
>>>> Anyway I have a few issues that I need clarifying:
>>>> 1 - Can you have too much overlap between windows?
>>> no, there no such a thing of too much overlap :)) You could even put
>>> two identical windows with same 100% overlap ... no problem.
>>>> 2 - Does the distribution at each window have to centered around the
>>>> desired r0? (If not does this affect the free energy?)
>>> The deviation of the distribution from the r0 is what dictates the
>>> profile. The more away from the disired r0 the higher the free energy
>>> of the system.
>>>> 3- If you over sample one particular window, will it affect the curve?
>>> There is no such a thing of over sampling ... the only thing you can
>>> have is not enough sampling.
>>>>
>>>> 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
>




More information about the gromacs.org_gmx-users mailing list