Hi David, Thanks for your suggestion to add coloring of isosurfaces based on gradient magnitude to Chimera. I think I can add this to the Surface Color tool in an hour or two since the gradient calculation code is already in Chimera (for fitting atomic models in maps). I'll try to put it in soon and send you the code to test. Description of coloring from article: "OpenDX (www.opendx.org) module MAKROVIS was implemented to create masks of different shapes and to analyze disorder in the reconstructions (Figure 1B). For this, isosurface rendering of a given voxel was colored as a function of the maximum density difference between the voxel and its neighbors." Tom
From: "David G. Morgan" <dmorgan@ucdavis.edu> To: goddard@cgl.ucsf.edu Subject: suggestion Date: Thu, 16 Nov 2006 19:03:50 -0800
Tom,
I have a suggestion for the developers "list of ideas to implement," if you think there would be enough interest from users and that it is something that chimera should do. I realize that you get lots of suggestions, and that some of them are simply outside the realm of what the chimera developers think chimera is intended to do. But since you do ask for suggestions, here's my most recent one.
I'm attaching a recent journal article where the authors used a color coding scheme to display "order/disorder" regions of an EM reconstruction. The images are in Fig 1 B and the rather sparse description of the "algorithm" is a single sentence in the "Image Processing 3D Reconstruction" section (p 1046). The idea of color coding some estimate of the better defined regions of a density map is very nice (and clever), and would be extremely helpful in lots of situations (my plug for wanting you guys to implement something similar...).
The idea behind this is that for any given region of iso-surface, the density gradient normal to that surface should be a reasonable measure of how well defined the density is. In other words, a very steep gradient (e.g., a fast transition between protein and solvent) means very well defined density and a flat gradient means weakly defined density (disordered regions, partial occupancy, etc.). The authors don't do such a complicated calculation, but it seems that what they implemented at least captures most of this sort of character.
I think the key to this is that the gradient is calculated "locally." I've looked at gradients as a function of resolution with the idea of using it to help chose a proper iso-surface. While it works well with model data, real reconstructions aren't nearly so well behaved. I hadn't thought about this in a while, but I think what the authors have done both helps explain the behavior of my real data (the gradient is simply too varied) and is a quite useful visualization tool.
So please think about implementing something like this in the volume handling tools of chimera.
David Gene Morgan Advanced Microscopy & Proteomics Molecular & Cellular Biology University of California at Davis 530 752 2693 (lab) 530 752 3085 (FAX) ---- removed PDF for
Huiskonen JT, de Haas F, Bubeck D, Bamford DH, Fuller SD, Butcher SJ. Structure of the bacteriophage phi6 nucleocapsid suggests a mechanism for sequential RNA packaging. Structure. 2006 Jun;14(6):1039-48.