aligning without preserving atom order

[Alexej Jerschow]

Is it possible to determine aligned RMSD and the rotation matrix without preserving the atom order? 

The situation I am considering is the following: 

I have a solvated ion in water (considering for example just the 1st salvation shell) and would like to determine if two different clusters have the same geometry (except for rotations), while also extracting the rotation matrix. In a normal alignment the order in which the atom positions are recorded matters for the alignment algorithm, but for this task it should specifically be disregarded. 

The documentation appears to include an option for such a task (for example the align_to functionalists has an option match_atoms, which appears to address this point, as it seems that it will just match by atom mass, but it is really unclear whether it would address the situation). Also, I did not see a provision for extracting the rotation matrix for this situation. 

Thank you

Alexej

[Oliver Beckstein]

Hi Alexej,

On Jul 23, 2020, at 3:30 PM, Alexej Jerschow <aj...@nyu.edu> wrote:

> Is it possible to determine aligned RMSD and the rotation matrix without preserving the atom order?

There’s no code in MDAnalysis that optimizes RMSD over different permutations of atoms.

>
> The situation I am considering is the following:
> I have a solvated ion in water (considering for example just the 1st salvation shell) and would like to determine if two different clusters have the same geometry (except for rotations), while also extracting the rotation matrix. In a normal alignment the order in which the atom positions are recorded matters for the alignment algorithm, but for this task it should specifically be disregarded.
>
> The documentation appears to include an option for such a task (for example the align_to functionalists has an option match_atoms, which appears to address this point, as it seems that it will just match by atom mass,

It won’t do a global matching and re-ordering.

There are ways with the select keyword to specify arbitrary selections for the mobile and reference atoms so *in principle* you could run it with many different permutations. But I doubt that this would give you reasonable performance and there are many permutations for water molecules...

> but it is really unclear whether it would address the situation). Also, I did not see a provision for extracting the rotation matrix for this situation.
>
> Thank you
> Alexej

This paper

F. Reinhard, O. Lange, J. Hub, J. Haas, and H. Grubmüller. g permute: Permutation-reduced phase space density compaction. Computer Physics Communications, 180(3):455–458, 2009. DOI: 10.1016/j.cpc.2008.10.018


might give you some ideas where to start if you want to implement something yourself.

Oliver

--
Oliver Beckstein
orbe...@gmail.com

[Alexej Jerschow]

Thank you very much for clarifying. Will check his out. 

Alexej