How does "super" compute its alignment score?

[danieljnilson]

As the title says.

I am writing a summary in my thesis of a pipeline I wrote that uses the PyMOL api. I'd like to be able to detail how the score is calculated as I make reference/use of it multiple times.

[ye11owSub]

hi @danieljnilson
Is this the code you were looking for?

[danieljnilson]

Hi @ye11owSub !
Thank you for your response. That is a good find; though it appears to be more of like the function super description. It does not seem to describe the generated output. I will try to use this information to dig more later.

[ye11owSub]

This function is written in C (Python is just an API for it), so it would be more helpful to look at this source code

[JarrettSJohnson]

The PyMOLWiki entry for super gives basic information https://pymolwiki.org/index.php/Super , and the C/C++ code is unfortunately not very clear to read. The major difference between this and the other align function, is that super first performs a sequence-independent residue pairing determined by local dihedral information (dihedral angles from CA-CB pairs along with other dihedrals against other atoms within a given radius (default 12 angstroms). Dynamic programming is used to match residue pairs by their local structural similarity; and from there, I believe you get your typical RMSD refinement stages just as you do with align.

Feel free to ask any questions if you need further clarification.

[danieljnilson]

Hello @JarrettSJohnson . Thank you for your response and the url link. I am familiar with that pymol wiki on super ; it is a very helpful page describing a high level overview of how that function works. Unfortunately, it does not describe how the "MatchAlign: score" is determined for each use of the super . I do not see a description in the wiki page for 'align' either. I will keep looking in the source code too. Thank you for your assistance.

[JarrettSJohnson]

From what I understand, the MatchAlign score takes an arbitrary scoring matrix along with dynamic programming parameters (gap penalties, etc...) to discover the optimal alignment via something similar to Needleman-Wunsch. For super, the scoring matrix is as I mentioned above, information regarding its local structural information (mostly dihedrals). While align will generate the scoring matrix based off of one like the BLOSUM62.

[danieljnilson]

Thank you. If there is a paper or some sort of reference to go with that, it would be great. However, for now this gives me something to work with.