[PyMOL] Ways to show overlapping peptides on structure

[Yang Su <su...@cr...>]

Hello,

I'm exploring ways to show multiple overlapping peptides on a protein
structure (see the picture below for an example). Being able to show
overlapping peptides all at once on structure is highly desirable in the
field of hydrogen-deuterium exchange mass spectrometry and may be useful in
other fields, too, but I don't know any structure visualization program
capable of doing that. I'll explain how I made the example picture and the
problems/pitfalls with that approach, and would like to hear your
suggestions for how to improve or alternative approaches.

Example picture:
https://pymolwiki.org/index.php/File:Overlapping_peptides.png
[image: email.png]

What I've done was to first "de-overlap" the peptides, i.e., to sort the
peptides into separate groups so that every group contains non-overlapping
peptides only. Each group of non-overlapping peptides was then shown in the
usual way on a "track". From pymol's perspective, a track is really just an
ordinary structure (only backbone atoms, no sidechains); the coordinates of
the backbone atoms are carefully offset from the real structure of the
protein, so that when all the tracks are displayed, they look like forming
an integral structure --- and this is where things get hacky:

1)  I need to know the orientation-/normal-/tangent-vectors that pymol uses
to draw secondary structure elements. Those vectors are calculated from
atomic coordinates in pymol. With them I can back-calculate the coordinates
of each track's atoms such that when pymol reads these coordinates and
displays as cartoon, the secondary structure elements align nicely across
all tracks and the final result looks like one integral structure. The
problem is pymol does not provide APIs to get those vectors, so I ended up
having to duplicate the relevant logic from pymol's source code
(layer2/RepCartoon.cpp) in my code to calculate those vectors. And that
made my code tied to a specific version of pymol, neither forward- nor
backward-compatible. Whenever pymol changes the way those vectors are
calculated, my code breaks.

2)  With how pymol calculates those vectors (~November 2018), there are
corner cases where the conversion between atomic coordinates and those
vectors are not easily reversible (for some beta-strands), which caused
ugly twists in some tracks.

3)  Because the atoms in the tracks are offset from the real structure in
different directions, by different amounts, to make the cartoon look nice,
the distances between bonding atoms often became longer, and I had to
increase 'connect_cutoff' to keep them connected, which sometimes screwed
up other parts of the structure.

4)  For large structures with many tracks, the number of atoms that pymol
needs to handle becomes huge, and the impact to performance and file size
is appreciable.

So, the above summarized my current effort, which mostly works, although
with its many pitfalls/drawbacks. As I said, I'm looking forward to
suggestions, alternatives, or any other ways to approach the central
problem: to display overlapping peptides on a structure.

Thank you,
Yang