From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Tue May 10 2011 - 10:43:51 CDT

On Tue, May 10, 2011 at 11:13 AM, Carolyn Phillips <phillicl_at_umich.edu> wrote:
> As a followup, if I reload the same point set and align all, it aligns perfectly and gives 0 RMSD.
>
> If I take the same point set, and exchange the index order of two of the points only, it aligns poorly, and gives a non-zero RMSD.
>
> So I think what I am learning here is that RMSD cannot "reregister" two sets of points, but can only best align the first to the first, the second to the second, and so forth.

carolyn,

right. that is the default of the underlying measure commands.

those _do_ support an "order" parameter, where you can map the order
of particles in one selection to the other, but that is not exposed to the GUI.

cheers,
      axel.

>
>
> On May 10, 2011, at 10:57 AM, Carolyn Phillips wrote:
>
>>
>> I have two sets of points defined on the surface of a unit  sphere that are nearly identical if rotated correctly.    I want to calculate the RMSD of these two sets of points.  I have converted them into a HOOMD XML file, in a fake box ten times the size of the particles,  and loaded them into VMD and view them as scaled down particles (Sphere Scale set to 0.2).  I am then using the RMSD GUI tool, but I am confused both by what I see on the screen and the output results.  I am not sure if this is because I am using RMSD for a non-bio molecule with very different units, or something else.
>>
>> So the first thing I notice is that the if I select index 0 for alignment,  the two sets of points appears in the viewer to be "almost" aligned uniformly.  However the two index 0's are not near each other at all.  Also, no single point is exactly aligned with the top index 0 point, but one appears nearly aligned.   However, the RMSD calculates to 0.   I am guessing that RMSD aligns by type, not index to index, is only calculating the top index 0 against the "nearest" particle of the same type,  has some kind of tolerance assuming the units of the system is "angstroms", and then I have no idea why the rest of the system is so well aligned.
>>
>> If I select index 1 for alignment, the two systems become dramatically not uniformly aligned, the RMSD calculates to 0, and again, the index 1 point is again visually not perfectly aligned with any other point.
>>
>> If select "all", I get something that looks pretty uniformly aligned.  My RMSD is not 0 (its about 0.88), and I assume it now represents the sum squared distance between all pairs of points in the two points sets, not just the closest pairs.
>>
>> What is the relationship between what I view and the actual alignment?    I drag or rotate independently one of the sets of the sets of point, the RMSD changes a tiny bit,  but not as if I am actually changing the alignment.  But If I align to a new atom selection, what is on the screen changes, but without appearing to perfectly align any two points.
>>
>> Thanks,
>> Carolyn
>>
>>
>>
>
>
>

-- 
Dr. Axel Kohlmeyer
akohlmey_at_gmail.com  http://goo.gl/1wk0
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.