cartoon tube "armadillos" in 3jac

[Elaine Meng]

In general cartoon tubes seem to work well, but I did find one structure that has some weird places that turn into "armadillos", 3jac (figure attached). Most of the tubes are fine, but a couple of them are armadillos.

In today's build with many structures I also get a lot of messages like these but it may just be from cartoon (not related to tubes) and usually the cartoons still look fine.

/Users/meng/Desktop/ChimeraX.app/Contents/Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/site-packages/chimerax/core/atomic/ribbon.py:891: RuntimeWarning: invalid value encountered in less
d[d < EPSILON] = 1

[Elaine Meng]

I see it is related to helices that are highly curved. There are 4 of these in 3jac and here I also list another example structure:

3JAC /a-c:2123-2149 /b:1130-1147
3X1L /a:513-534 /h:135-163

I was hoping that in the files they were actually defined as more than one helix, but at least in these cases they were defined as a single helix. However, running "dssp" actually made the armadillos worse in both structures... in 3X1L it even created 3 more armadillos. Just out of curiosity I tried Chimera pipes-n-planks on 3JAC, but it wasn't good: if helix splitting was false, it just put one straight pipe through the U-shaped helix, and if helix splitting was true, it made lots of very very short pipes.

Meanwhile testing these structures I get many, many more of the EPSILON warning mentioned in the initial report. However, I also get this warning when just showing default ribbons for some structures that seem to look OK (e.g. 4Y7K).

[Conrad Huang]

In 3jac, residues /b:2112-2148 should really be three short helices. However, dssp classifies them as one long helix. As a tube, it wraps around the arc approximately 270 degrees, which violated my assumption that helical arcs are not "too curved". The code was updated 2ea5536 to sort the order of tube centers differently and to cap the maximum tube radius to 2.5A. That range of residues now looks donut-like (with a bite taken out); it probably should look like a triangle if you look at the ribbon representation of it, but that won't happen until we put in code to analyze individual helix geometries.