Re: Issues with 'Gyration' in Colvars module

From: Aron Broom (broomsday_at_gmail.com)
Date: Wed May 23 2012 - 18:09:29 CDT

Hi Shaon,

>From my limited knowledge, your files look reasonable (but Giacomo is the
expert). What happens if you run the simulation with ABF but with the
biasing turned off? Would the Rg also tend towards ~4? If so, and 4 is
the local energy minimum (or even global) for the Rg, maybe you just need
stronger walls? I often use wall constants in the range that you are using
for the distance of a ligand, but if your entire molecular wants to
collapse, it could be that the potential energy for that collapse is just
greater than the wall.

Also, one thing that might help is to turn on "outputappliedforces" (the
spelling may be different) for your Rg colvar, so that you can more
effectively see what (if anything) the ABF is actually doing to your
simulation, particularly as it nears/crosses the wall.

Along the same lines as above, the Rg seems to cross the wall quite
quickly, based on your trajectory is moves across in less than 1000 steps,
and given that your fullsamples is 1000, that means that it moves through
the wall before any biasing is even taking place, so it must be entirely an
effect of the unbiased free energy landscape of Rg, again suggesting that
you may just need bigger walls.

Are you doing constant pressure and temperature? And are these the same
between the equilibration run you did and the ABF run? Also, if you are
doing constant pressure, did you use the *.xsc file for the extended
system? I can imagine any number of issues with pressure and periodicity
that might be causing your system to get crushed, thereby forcing
unreasonable collapse of your molecule.

~Aron

On Wed, May 23, 2012 at 6:25 PM, Shaon Chakrabarti <shaonc_at_gmail.com> wrote:

>
> ....And here are the details of 1)
>
> The following calculations have been
> done on a pdb file (which I could send
> separately) grabbed from the end of
> a trajectory run. The psf file I used in
> VMD is the same one I used to run
> the NAMD simulation.
>
> MY CALCULATION: (x,y,z denote atom coordinates)
> a) Rg defined with center of mass:
> Rg = sqrt[1/N * sum over all N atoms{(x-COMx)^2+(y-COMy)^2+(z-COMz)^2} ]
>
> With this definition I get Rg = 5.7555
>
> b) Rg defined with center of geometry:
> COGx = (1/N * sum of all x components). Similarly for COGy, COGz
> Rg = sqrt[1/N * sum over all N atoms{(x-COGx)^2+(y-COGy)^2+(z-COGz)^2}
> ]
>
> With this definition I get Rg = 5.754163
>
> VMD MEASURE RGYR:
> this gives Rg = 5.7541623 --almost identical to the result of part b)
> above
>
> COLVARS OUTPUT (from colvars.traj file):
> this gives Rg = 5.579920
>
> Hope this helps...
>
> Thanks,
> Shaon
>
>
>
>
>
> On Wed, May 23, 2012 at 3:11 PM, Shaon Chakrabarti <shaonc_at_gmail.com>wrote:
>
>> Hi Aron and Giacomo,
>> Thanks a lot for such quick replies! Actually I had
>> already done what Aron suggested: I first ran a 6ns
>> long equilibration trajectory without the ABF module,
>> but with gyration on.
>> From the last parts of this trajectory, I created pdb
>> and psf files of the molecule with Rg within the
>> required range. Then I reinitialized the velocities and
>> ran the trajectory with ABF included. I'm attaching the
>> initial bit of colvars.traj file here--the range is 7-11 (i had the
>> file for this range). The starting Rg value is about 8...
>> so that is between 7 and 11...yet, the Rg value doesn't
>> stay within the required range!
>> I'm also attaching the colvars configuration file if that helps.
>>
>> About 1) ...I will provide the numbers and details of my code in
>> a little while.
>>
>> Thanks again for all the help,
>>
>> Shaon
>>
>>
>>
>>
>>
>>
>>
>> On Wed, May 23, 2012 at 1:27 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com
>> > wrote:
>>
>>> Hi Shaon, Aron has already answered well on 2).
>>>
>>> About 1), can you provide the exact numbers, and more details regarding
>>> your own calculation? Do you load the same PSF topology and coordinates in
>>> VMD or use the masses guessed by VMD?
>>>
>>> In the colvars module, gyration is computed as the mass-weighted square
>>> distance from the center of geometry, taken under square root. But I
>>> actually realize now that the formula in the user's guide:
>>>
>>> http://www.ks.uiuc.edu/Research/namd/2.9/ug/node55.html#SECTION0001322170000000000000
>>> does not include the weighting masses. Would that explain the small
>>> difference that you saw? If so, I apologize for the omission of the masses
>>> in the user's guide.
>>>
>>> The gyration defined in VMD is also mass-weighted, but uses the center
>>> of mass.
>>>
>>> In gyration, it seems contradictory to have mass-weighted gyration
>>> around the center of geometry and not mass, but we decided towards that
>>> considered that gyration can be and is often used in combination with RMSD.
>>>
>>> On Wed, May 23, 2012 at 3:56 PM, Aron Broom <broomsday_at_gmail.com> wrote:
>>>
>>>> Hi Shaon,
>>>>
>>>> I've no idea about 1), but for 2), what happens if you specifiy Rg to
>>>> be within 4-7 rather than 5-7? From my understanding the ABF method needs
>>>> to sample forces before it starts to apply a bias, so if you're system
>>>> starts outside the biasing range, it will never collect any samples, and
>>>> hence never start to apply a bias. That would be my first thought anyway.
>>>> If you really want to do 5-7 angstroms, you should apply a harmonic
>>>> potential on Rg and sit it ~6 angstroms in order to equilibrate the system
>>>> before running ABF.
>>>>
>>>> ~Aron
>>>>
>>>>
>>>> On Wed, May 23, 2012 at 3:48 PM, Shaon Chakrabarti <shaonc_at_gmail.com>wrote:
>>>>
>>>>> Hi NAMD experts,
>>>>> I'm facing two problems with the gyration component in
>>>>> the Colvars module.
>>>>>
>>>>> 1) The NAMD user's guide says that the radius of gyration
>>>>> is defined using the Center of Geometry. However, when
>>>>> I write my own code (using the COG definition) to compute
>>>>> the Rg of a molecule, the Rg value is always about 0.2
>>>>> angstroms more compared to the output in the 'colvars.traj'
>>>>> file. The values from my code exactly match the values
>>>>> I get if I use the 'measure rgyr' command in VMD. I also
>>>>> computed Rg using Center of Mass, but that makes the
>>>>> difference even larger. I'd be grateful if someone can tell
>>>>> me if I'm making a mistake somewhere, or if Rg is
>>>>> actually defined differently in the Colvars module.
>>>>> P.S. My molecule (about 30 angstroms long when fully
>>>>> extended) is in the middle of a water box of side 55
>>>>> angstroms.
>>>>>
>>>>>
>>>>>
>>>>> 2) The second problem is similar (as I understand from
>>>>> the threads in 2008) to what Subramanian
>>>>> Vaitheeswaran faced while using the gyration
>>>>> component with ABF. If I specify the Rg to be (say)
>>>>> within say 5 and 7 angstroms, and use ABF to get
>>>>> the PMF, the system hovers at Rg value around
>>>>> 4.5-4.8 most of the time! A final resolution doesn't
>>>>> seem to have been posted back in 2008, and it
>>>>> would be of immense help if someone could tell
>>>>> me what's going on....if the problem has been
>>>>> resolved. I tried switching the hideJacobian option
>>>>> to yes, and increasing the lower and upper wall
>>>>> constants, but these don't make any difference.
>>>>>
>>>>>
>>>>> It would be great if someone could help with these two issues.
>>>>>
>>>>> Thanks,
>>>>> Shaon
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>
>>>>
>>>> --
>>>> Aron Broom M.Sc
>>>> PhD Student
>>>> Department of Chemistry
>>>> University of Waterloo
>>>>
>>>>
>>>
>>
>

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

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:32 CST