[Molvis-list] FirstGlance /Jmol: A few observations

[Gale Rhodes]

On Feb 13, 2006, at 1:53 PM, Eric Martz wrote:

> A new version of FirstGlance in Jmol (http://firstglance.jmol.org)  
> has been released.

Great job, Eric!

A couple of things about Jmol itself:

1) Is it just me, or are Jmol's new (v. 10.00.46) stereographic  
options backward? I'm referring to [Render:Stereographic:Cross-eyed  
(or Wall-eyed) viewing.] I believe that the current Jmol has them  
reversed. I get cross-eye stereo (convergent) when I ask for wall-eye  
(divergent) and vice versa.

If I use the console to turn on stereo, the command "stereo 4" gives  
wall-eye, just as I expect from RasMol or DeepView. And the console  
command "stereo -4" gives cross-eye. But the menu commands don't seem  
to be doing the same thing.

Any other stereo users out there want to check? Stereo is only  
available if the viewer has been upgraded to Jmol 10.00.46.  
Firstglance and the samples at Jmol home are upgraded. PDB's Jmol  
viewer and the ProteinGlimpse Widget are not.

2) One can always think of something to add to these wonderful tools.  
It occurred to me, as I was preparing a talk on judging model  
quality, that Jmol has no built-in options for helping model users to  
validate models.
	a) Color by B-factor should be a simple addition, and B-factors can  
alert you to certain kinds of model problems. With the same command,  
NMR models are colored by how much each atom's position varies across  
the ensemble of models (rmsd from the average position), also a  
quality indicator. (The same PDB file column that contains B-factors  
for x-ray models contains rmsds for NMR models, and contains  
uncertainty indicators for homology models, so one command fits all.)
	b) A more useful tool in model validation is a Ramachandran diagram,  
which can expose even some rather subtle model errors, and can be  
useful in sorting out alternative templates for homology modeling.  
Modelers avoid templates that have a lot of Ramachandran outliers.  
The resulting model will have the same outliers.
	c) Other useful tools would be "Select:Clashing Residues" and  
"Select:Residues with unsatisfied H-bonds." The latter finds buried  
polar side chains that are probably in the wrong place.

Such errors are present in many PDB files, and occur commonly in  
routine modeling from sequence data using PDB models as templates.  
Viewers that do not have model validation options make all models  
look good, and that's not the truth.

3) Finally, thanks to Jmol developers for giving what looks like a  
solid foundation for many powerful tools for molecular viewing and  
analysis.

Cheers!

Gale Rhodes, Professor of Chemistry
University of Southern Maine
PO Box 9300
Portland, Maine 04104-9300
http://www.usm.maine.edu/~rhodes